Publish updated SamAI Tokenizer repository

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+assets/SamAI-Tokenizer-Benchmark.png filter=lfs diff=lfs merge=lfs -text

CITATION.cff

@@ -1,6 +1,6 @@
 cff-version: 1.2.0
-message: "If you use SunAI-64K, please cite this repository."
-title: "SunAI-64K: An Arabic-Focused Byte-Level BPE Tokenizer"
+message: "If you use SamAI-Tokenizer, please cite this repository."
+title: "SamAI-Tokenizer: An Arabic-Focused Byte-Level BPE Tokenizer"
 type: software
 authors:
   - family-names: Aljohani
@@ -8,3 +8,4 @@ authors:
 version: 1.0.0
 date-released: 2026-09-20
 license: Apache-2.0
+repository-code: "https://huggingface.co/Samai-sa/SamAI-Tokenizer"

NOTICE

@@ -1,4 +1,4 @@
-SunAI-64K
+SamAI-Tokenizer
 Copyright 2026 Yasser Aljohani
 
 This product includes aggregate benchmark results produced from third-party

README.md

@@ -1,25 +1,36 @@
 ---
 language:
-- ar
-- en
+  - ar
+  - en
 license: apache-2.0
 library_name: tokenizers
+datasets:
+  - AdaMLLab/AraMix
+  - HuggingFaceFW/fineweb-2
+  - HuggingFaceFW/fineweb
 tags:
-- tokenizer
-- arabic
-- arabic-nlp
+  - tokenizer
+  - arabic
+  - arabic-nlp
+  - bpe
+  - byte-level
+  - multilingual
+  - saudi-arabia
 ---
 
 # SamAI-Tokenizer
 
-SamAI-Tokenizer is an Arabic-focused, byte-level BPE tokenizer intended for training
-new Arabic-heavy language models. It has a vocabulary of 65,536 tokens, uses
-NFC normalization, preserves the original text during decoding, and includes
-reserved tokens for future model development.
+[**🚀 Try the interactive tokenizer demo**](https://huggingface.co/spaces/Samai-sa/SamAI-Tokenizer-Demo) · [**العربية**](README_AR.md)
 
-This repository contains the complete tokenizer, validation reports, the
-four-dataset benchmark implementation, and aggregate benchmark results. It
-does **not** contain model weights or redistribute training/benchmark text.
+SamAI-Tokenizer is an Arabic-focused byte-level BPE tokenizer for training new
+Arabic-heavy language models. It has a vocabulary of 65,536 tokens, applies
+Unicode NFC normalization, supports lossless decoding, and retains English
+coverage for multilingual training.
+
+This repository contains the complete tokenizer, Hugging Face configuration,
+usage examples, validation reports, reproducible benchmark code, and aggregate
+results. It does **not** contain model weights, training-corpus text, or
+benchmark documents.
 
 ## Key specifications
 
@@ -32,6 +43,8 @@ does **not** contain model weights or redistribute training/benchmark text.
 | Unknown-token fallback | Byte-level coverage |
 | Named special tokens | `<pad>`, `<bos>`, `<eos>`, `<unk>`, `<user>`, `<assistant>`, `<system>` |
 | Reserved token IDs | 7–255 |
+| Primary language | Arabic |
+| Secondary language | English |
 
 ## Installation
 
@@ -41,8 +54,6 @@ pip install -U tokenizers transformers huggingface_hub
 
 ## Usage
 
-Replace `Samai-sa/SamAI-Tokenizer` with the published repository ID.
-
 ### Transformers
 
 ```python
@@ -55,9 +66,10 @@ tokenizer = PreTrainedTokenizerFast.from_pretrained(
 text = "المملكة العربية السعودية"
 ids = tokenizer.encode(text, add_special_tokens=False)
 
-print(ids)
-print(tokenizer.convert_ids_to_tokens(ids))
-print(tokenizer.decode(ids))
+print("Token count:", len(ids))
+print("Token IDs:", ids)
+print("Tokens:", tokenizer.convert_ids_to_tokens(ids))
+print("Decoded:", tokenizer.decode(ids))
 ```
 
 ### Tokenizers
@@ -77,15 +89,26 @@ encoding = tokenizer.encode(
     add_special_tokens=False,
 )
 
-print(encoding.ids)
-print(encoding.tokens)
-print(tokenizer.decode(encoding.ids))
+print("Token count:", len(encoding.ids))
+print("Token IDs:", encoding.ids)
+print("Tokens:", encoding.tokens)
+print("Decoded:", tokenizer.decode(encoding.ids))
 ```
 
+## Interactive demo
+
+The [SamAI Tokenizer Playground](https://huggingface.co/spaces/Samai-sa/SamAI-Tokenizer-Demo)
+lets users enter Arabic or English text and inspect:
+
+- token count and tokens per word;
+- token IDs and raw token pieces;
+- original text spans and offsets;
+- decoded text for round-trip verification.
+
 ## Training mixture
 
-The tokenizer was trained from sampled text covering Arabic and English. The
-training text itself is not included.
+The tokenizer was trained on sampled Arabic and English text. The processed
+training text is not distributed in this repository.
 
 | Source bucket | Characters used |
 | --- | ---: |
@@ -95,30 +118,35 @@ training text itself is not included.
 | FineWeb English | 200,000,000 |
 | **Total** | **629,856,611** |
 
-
 ## Validation
 
-The supplied validation run reported zero round-trip failures and zero unknown
-tokens across Saudi schoolbooks, AraMix Arabic, FineWeb2 Arabic, and FineWeb
-English samples. See [`reports/validation.json`](reports/validation.json) and
+The released validation run reported zero round-trip failures and zero unknown
+tokens across samples from Saudi schoolbooks, AraMix Arabic, FineWeb2 Arabic,
+and FineWeb English. Additional stress tests cover Modern Standard Arabic,
+dialects, Arabizi, English, code, mathematics, and Unicode edge cases.
+
+See [`reports/validation.json`](reports/validation.json) and
 [`reports/stress.json`](reports/stress.json).
 
-## Four-dataset Arabic benchmark
+## Arabic tokenizer benchmark
 
-SamAI-Tokenizer was compared on four complementary Arabic datasets:
+SamAI-Tokenizer was evaluated on four complementary Arabic benchmarks. Lower
+tokens per word indicates more efficient tokenization.
 
-| Benchmark | Register | Documents | Words |
-| --- | --- | ---: | ---: |
-| ArabicMMLU | Academic questions and reading passages | 14,435 | 410,566 |
-| MLQA Arabic | Questions and long answer-bearing passages | 10,933 | 638,708 |
-| MASSIVE Arabic | Short assistant commands | 2,907 | 15,356 |
-| Arabic Dialects Dataset | Regional dialect text | 9,807 | 337,653 |
-| **Combined** |  | **38,082** | **1,402,283** |
+![SamAI tokenizer benchmark results](assets/SamAI-Tokenizer-Benchmark.png)
 
-The primary ranking is an equal-weight macro average: each benchmark receives
-25% weight regardless of corpus size. Lower tokens per word is better.
+| Benchmark | SamAI tokens/word | Best tokenizer | Best tokens/word | SamAI rank |
+| --- | ---: | --- | ---: | ---: |
+| Arabic Dialects | **1.3043** | ALLaM-7B | 1.2812 | **2** |
+| ArabicMMLU | **2.4838** | ALLaM-7B | 1.9943 | **5** |
+| MASSIVE Arabic | **1.2549** | ALLaM-7B | 1.1939 | **2** |
+| MLQA Arabic | **1.4397** | **SamAI-Tokenizer** | **1.4397** | **1** |
 
-| Rank | Tokenizer | Mean tokens/word | Tokens versus SunAI |
+Across the equal-weight macro average of the four benchmarks, SamAI-Tokenizer
+ranked second among 15 successfully tested tokenizers at **1.6207 tokens per
+word**.
+
+| Rank | Tokenizer | Mean tokens/word | Extra tokens versus SamAI |
 | ---: | --- | ---: | ---: |
 | 1 | ALLaM-7B | 1.5021 | -4.86% |
 | **2** | **SamAI-Tokenizer** | **1.6207** | **baseline** |
@@ -136,37 +164,19 @@ The primary ranking is an equal-weight macro average: each benchmark receives
 | 14 | Kimi-K2.6 | 2.8693 | +84.13% |
 | 15 | tiktoken-cl100k | 4.0800 | +164.63% |
 
-### Results by benchmark
-
-| Benchmark | SunAI tokens/word | Best tokenizer | Best tokens/word | SunAI position |
-| --- | ---: | --- | ---: | ---: |
-| Arabic Dialects | 1.3043 | ALLaM-7B | 1.2812 | 2 |
-| ArabicMMLU | 2.4838 | ALLaM-7B | 1.9943 | 5 |
-| MASSIVE Arabic | 1.2549 | ALLaM-7B | 1.1939 | 2 |
-| MLQA Arabic | **1.4397** | **SamAI-Tokenizer** | **1.4397** | **1** |
-
-Complete aggregate results are under [`results/`](results/). The row-level
-benchmark output is intentionally excluded because it reproduces third-party
-dataset text; the script can regenerate it for users who accept the applicable
-dataset terms.
+Complete aggregate results are available in [`results/`](results/). Row-level
+outputs are intentionally excluded because they reproduce third-party text.
 
 ## Benchmark methodology
 
 - All available examples were used, followed by exact-text deduplication within
   each benchmark.
 - Token counts exclude chat templates and added special tokens.
-- `tokens_per_word` is total tokens divided by whitespace-separated words.
-- The macro ranking averages each benchmark's ratio to SunAI, giving every
-  benchmark equal weight.
-- The pooled report instead sums all tokens and words before calculating the
-  ratio; MLQA and ArabicMMLU therefore receive more weight in that view.
-- Qwen2.5 and Qwen3 produced identical results. Qwen3.5 and Qwen3.8 also
-  produced identical results. GPT-OSS and `o200k_base` produced identical
-  results. These should be interpreted as tokenizer-family duplicates, not
+- Tokens per word is total tokens divided by whitespace-separated words.
+- The primary macro ranking gives each benchmark 25% weight.
+- The pooled report sums all tokens and words before calculating the ratio.
+- Tokenizers producing identical results are tokenizer-family duplicates, not
   independent confirmations.
-- Jais-13B, Gemma-3-4B, Llama-4-Scout, and Aya-Expanse were gated during the
-  published run. AceGPT-v2 could not be loaded by the tested library version.
-  They are not included in the ranking.
 
 To reproduce the benchmark:
 
@@ -175,39 +185,53 @@ pip install -r benchmark/requirements.txt
 python benchmark/benchmark.py
 ```
 
+Third-party remote code is disabled by default. Review and pin relevant
+repositories before opting in:
+
+```bash
+SAMAI_TRUST_REMOTE_CODE=1 python benchmark/benchmark.py
+```
 
 ## Limitations
 
-- Tokenization efficiency does not measure language-model quality, factuality,
-  safety, reasoning, or downstream accuracy.
+- Tokenization efficiency does not measure language-model quality, reasoning,
+  factuality, safety, or downstream accuracy.
 - Fully vocalized Arabic is the clearest weakness. One heavily diacritized
-  ArabicMMLU subject accounts for most of the gap to ALLaM. Excluding `Arabic
-  Language (General)`, SunAI records approximately 1.630 tokens/word versus
-  ALLaM's 1.709 on the remaining ArabicMMLU subjects.
-- ArabicMMLU contains reading passages repeated across different questions.
-  Exact full-document deduplication does not remove every repeated passage.
-- Whitespace-separated words are a transparent efficiency denominator, but
-  they are not a linguistic segmentation of Arabic morphology.
-- The tokenizer was designed for training new models. Replacing the tokenizer
-  of an already pretrained model requires resizing and retraining its embedding
-  and output layers; it is not a drop-in vocabulary replacement.
+  ArabicMMLU category accounts for most of the gap to ALLaM.
+- Whitespace-separated words are a transparent denominator, not a linguistic
+  segmentation of Arabic morphology.
+- Replacing the tokenizer of an already pretrained model requires resizing and
+  retraining its embedding and output layers. This is not a drop-in vocabulary
+  replacement.
+
+## Repository contents
+
+| Path | Description |
+| --- | --- |
+| `tokenizer.json` | Complete tokenizer |
+| `tokenizer_config.json` | Transformers tokenizer configuration |
+| `special_tokens_map.json` | Named special-token mapping |
+| `example.py` | Minimal loading and token-count example |
+| `validate_tokenizer.py` | Vocabulary and round-trip validation |
+| `benchmark/` | Reproducible four-benchmark implementation |
+| `results/` | Published aggregate benchmark results |
+| `reports/` | Validation and stress-test reports |
+| `assets/` | Model-card benchmark visualization |
 
 ## License and data terms
 
 The tokenizer files and repository code are released under Apache-2.0. Source
-datasets and benchmark datasets remain subject to their own licenses and terms.
-No source-corpus or benchmark documents are distributed here. Users are
-responsible for confirming that their intended use complies with the relevant
-upstream terms.
+and benchmark datasets remain subject to their own licenses and terms. No
+source-corpus or benchmark documents are redistributed here.
 
 ## Citation
 
 ```bibtex
-@software{aljohani2026sunai64k,
+@software{aljohani2026samaitokenizer,
   author  = {Yasser Aljohani},
   title   = {SamAI-Tokenizer: An Arabic-Focused Byte-Level BPE Tokenizer},
   year    = {2026},
   version = {1.0.0},
-  url     = {https://huggingface.co/SamAI/SamAI-Tokenizer}
+  url     = {https://huggingface.co/Samai-sa/SamAI-Tokenizer}
 }
-```
+```

README_AR.md

@@ -1,23 +1,26 @@
 # SamAI-Tokenizer
 
-`SamAI-Tokenizer` هو مقطّع نصوص عربي مبني بخوارزمية BPE على مستوى البايت، ومصمم
-لتدريب نماذج لغوية جديدة تركّز على اللغة العربية. يبلغ حجم القاموس 65,536
-رمزًا، ويستخدم Unicode NFC مع إمكانية استعادة النص الأصلي بعد فك الترميز.
+[**🚀 جرّب المقطّع تفاعليًا**](https://huggingface.co/spaces/Samai-sa/SamAI-Tokenizer-Demo) · [**English**](README.md)
 
-تتضمن الحزمة ملف `tokenizer.json` الكامل، وملفات إعداد Hugging Face، وتقارير
-التحقق، ونتائج مقارنة على أربع مجموعات بيانات عربية، إضافة إلى سكربت إعادة
-تشغيل المقارنة.
+`SamAI-Tokenizer` مقطّع نصوص يركّز على اللغة العربية، ومبني بخوارزمية BPE
+على مستوى البايت. صُمم لتدريب نماذج لغوية جديدة ذات محتوى عربي كثيف، ويحتوي
+على قاموس من 65,536 رمزًا مع تطبيع Unicode NFC وإمكانية استعادة النص بعد فك
+الترميز.
 
-## أبرز النتائج
+تحتوي هذه الحزمة على المقطّع كاملًا، وإعدادات Hugging Face، وأمثلة الاستخدام،
+وتقارير التحقق، وكود الاختبارات، والنتائج التجميعية. ولا تحتوي على أوزان نموذج
+لغوي أو نصوص بيانات التدريب والاختبارات.
 
-- المركز الثاني في المتوسط المتساوي لأربعة اختبارات من بين 15 مقطّعًا تم
-  تشغيلها بنجاح.
-- 1.6207 رمز لكل كلمة في المتوسط.
-- المركز الأول على MLQA Arabic بمعدل 1.4397 رمز لكل كلمة.
-- يتفوق في كفاءة التقطيع على Fanar-1 بنسبة 36.81%، وعلى `o200k_base` بنسبة
-  26.37%، وعلى `cl100k_base` بنسبة 164.63% وفق المتوسط المتساوي.
-- الضعف الأوضح هو النص العربي المشكول بالكامل، ويجب أخذه في الاعتبار عند
-  تدريب نماذج للقرآن أو النصوص التراثية والتعليمية المشكولة.
+## المواصفات
+
+| الخاصية | القيمة |
+| --- | --- |
+| الخوارزمية | Byte-level BPE |
+| حجم القاموس | 65,536 |
+| التطبيع | Unicode NFC |
+| اللغة الأساسية | العربية |
+| اللغة الثانوية | الإنجليزية |
+| الرموز الخاصة | `<pad>`, `<bos>`, `<eos>`, `<unk>`, `<user>`, `<assistant>`, `<system>` |
 
 ## الاستخدام
 
@@ -25,15 +28,63 @@
 from transformers import PreTrainedTokenizerFast
 
 tokenizer = PreTrainedTokenizerFast.from_pretrained(
-    "SamAI/SamAI-Tokenizer"
+    "Samai-sa/SamAI-Tokenizer"
 )
 
 text = "المملكة العربية السعودية"
 ids = tokenizer.encode(text, add_special_tokens=False)
 
-print(ids)
-print(tokenizer.convert_ids_to_tokens(ids))
-print(tokenizer.decode(ids))
+print("عدد الرموز:", len(ids))
+print("معرّفات الرموز:", ids)
+print("النص بعد فك الترميز:", tokenizer.decode(ids))
 ```
 
-التفاصيل الكاملة والمنهجية والقيود موجودة في [`README.md`](README.md).
+## العرض التفاعلي
+
+يتيح [SamAI Tokenizer Playground](https://huggingface.co/spaces/Samai-sa/SamAI-Tokenizer-Demo)
+إدخال نص عربي أو إنجليزي وعرض عدد الرموز، والرموز لكل كلمة، ومعرّفات الرموز،
+والمقاطع النصية، والإزاحات، والنص بعد فك الترميز.
+
+## بيانات التدريب
+
+| المصدر | عدد المحارف المستخدمة |
+| --- | ---: |
+| الكتب المدرسية السعودية | 29,856,611 |
+| AraMix Arabic | 200,000,000 |
+| FineWeb2 Arabic | 200,000,000 |
+| FineWeb English | 200,000,000 |
+| **الإجمالي** | **629,856,611** |
+
+## نتائج الاختبارات
+
+كلما انخفض عدد الرموز لكل كلمة كانت كفاءة التقطيع أفضل.
+
+![نتائج مقارنة SamAI-Tokenizer](assets/SamAI-Tokenizer-Benchmark.png)
+
+| الاختبار | SamAI رمز/كلمة | ترتيب SamAI |
+| --- | ---: | ---: |
+| Arabic Dialects | 1.3043 | **2** |
+| ArabicMMLU | 2.4838 | **5** |
+| MASSIVE Arabic | 1.2549 | **2** |
+| MLQA Arabic | 1.4397 | **1** |
+
+حقق `SamAI-Tokenizer` المركز الثاني من بين 15 مقطّعًا تم تشغيلها بنجاح في
+المتوسط المتساوي للاختبارات الأربعة، بمعدل **1.6207 رمز لكل كلمة**.
+
+النتائج التفصيلية موجودة في مجلد [`results/`](results/)، ومنهجية إعادة التشغيل
+موجودة في [`benchmark/README.md`](benchmark/README.md).
+
+## التحقق والقيود
+
+- لم ت��جل عينات التحقق أي فشل في استعادة النص أو رموزًا مجهولة.
+- كفاءة التقطيع لا تقيس جودة النموذج اللغوي أو قدرته على الاستدلال.
+- النص العربي المشكول بالكامل هو أبرز نقاط الضعف الحالية.
+- لا يمكن استبدال مقطّع نموذج مدرّب مسبقًا مباشرة دون إعادة تدريب طبقات
+  التضمين والإخراج.
+
+## الترخيص
+
+ملفات المقطّع وكود المستودع متاحة بترخيص Apache-2.0. وتبقى بيانات المصدر
+والاختبارات خاضعة لتراخيصها وشروطها الأصلية.
+
+للتفاصيل الكاملة راجع [`README.md`](README.md).

benchmark/README.md

@@ -1,6 +1,6 @@
 # Benchmark reproduction
 
-The benchmark compares SunAI-64K with Arabic-focused and recent multilingual
+The benchmark compares SamAI-Tokenizer with Arabic-focused and recent multilingual
 tokenizers on four datasets. It downloads tokenizer assets only, not model
 weights.
 
@@ -19,8 +19,8 @@ Generated files are written to `results/generated/`. Override paths when
 needed:
 
 ```bash
-SUNAI_TOKENIZER_PATH=/path/to/tokenizer.json \
-SUNAI_BENCHMARK_OUTPUT=/path/to/results \
+SAMAI_TOKENIZER_PATH=/path/to/tokenizer.json \
+SAMAI_BENCHMARK_OUTPUT=/path/to/results \
 python benchmark/benchmark.py
 ```
 
@@ -28,7 +28,7 @@ Third-party remote code is disabled by default. Some custom tokenizers will be
 skipped unless you review their repositories and explicitly opt in:
 
 ```bash
-SUNAI_TRUST_REMOTE_CODE=1 python benchmark/benchmark.py
+SAMAI_TRUST_REMOTE_CODE=1 python benchmark/benchmark.py
 ```
 
 Gated repositories also require accepting their terms and authenticating with

benchmark/benchmark.py

@@ -18,7 +18,7 @@ from transformers import AutoTokenizer
 REPO_ROOT = Path(__file__).resolve().parents[1]
 
 MY_TOKENIZER_PATH = Path(
-    os.environ.get("SUNAI_TOKENIZER_PATH", REPO_ROOT / "tokenizer.json")
+    os.environ.get("SAMAI_TOKENIZER_PATH", REPO_ROOT / "tokenizer.json")
 )
 
 # Four complementary Arabic benchmarks. Each loader uses evaluation/test data
@@ -39,7 +39,7 @@ RANDOM_SEED = 42
 
 OUTPUT_DIR = Path(
     os.environ.get(
-        "SUNAI_BENCHMARK_OUTPUT",
+        "SAMAI_BENCHMARK_OUTPUT",
         REPO_ROOT / "results" / "generated",
     )
 )
@@ -47,9 +47,9 @@ OUTPUT_DIR = Path(
 # Some third-party tokenizers require executing code from their repositories.
 # Keep this disabled by default. Review those repositories and enable it only
 # when you explicitly accept that risk:
-#   SUNAI_TRUST_REMOTE_CODE=1 python benchmark/benchmark.py
+#   SAMAI_TRUST_REMOTE_CODE=1 python benchmark/benchmark.py
 TRUST_REMOTE_CODE = os.environ.get(
-    "SUNAI_TRUST_REMOTE_CODE", "0"
+    "SAMAI_TRUST_REMOTE_CODE", "0"
 ).lower() in {"1", "true", "yes"}
 
 # Only tokenizer/configuration files are downloaded, not model weights.
@@ -483,7 +483,7 @@ def build_summary(frame, token_columns, benchmark_name):
     total_characters = int(frame["characters"].sum())
     total_bytes = int(frame["bytes"].sum())
 
-    custom_total = int(frame[token_columns["SunAI-64K"]].sum())
+    custom_total = int(frame[token_columns["SamAI-Tokenizer"]].sum())
     rows = []
 
     for tokenizer_name, column in token_columns.items():
@@ -537,7 +537,7 @@ def build_subject_report(frame, token_columns):
     ):
         words = int(subset["words"].sum())
         byte_count = int(subset["bytes"].sum())
-        custom_tokens = int(subset[token_columns["SunAI-64K"]].sum())
+        custom_tokens = int(subset[token_columns["SamAI-Tokenizer"]].sum())
 
         for tokenizer_name, column in token_columns.items():
             tokens = int(subset[column].sum())
@@ -573,9 +573,9 @@ def main():
     token_columns = {}
 
     # Your tokenizer
-    column = "tokens__SunAI_64K"
+    column = "tokens__SamAI_Tokenizer"
     frame[column] = benchmark_custom_tokenizer(frame)
-    token_columns["SunAI-64K"] = column
+    token_columns["SamAI-Tokenizer"] = column
 
     # Hugging Face competitors
     for name, repository in HF_TOKENIZERS.items():

example.py

@@ -1,8 +1,7 @@
 from transformers import PreTrainedTokenizerFast
 
 
-# Replace this with your published Hugging Face repository ID.
-REPO_ID = "SamAI/SamAI-Tokenizer"
+REPO_ID = "Samai-sa/SamAI-Tokenizer"
 
 tokenizer = PreTrainedTokenizerFast.from_pretrained(REPO_ID)
 

meta.json

@@ -10,6 +10,5 @@
     "fineweb2_arabic": 200000000,
     "fineweb_english": 200000000
   },
-  "training_sha256": "bf02ef3cfa689f435aa5402ace1cb09221f9bec22b1f19dd6e73eaf1bddbd115",
   "normalization": "NFC"
 }

reports/stress.json

@@ -1,5 +1,4 @@
 {
-  "tokenizer_sha256": "7d9496a251b44688557a79fab59b5cf7a12f11a9559540b9a1e223657f35f613",
   "groups": {
     "msa": {
       "documents": 1,

reports/validation.json

@@ -1,5 +1,4 @@
 {
-  "tokenizer_sha256": "7d9496a251b44688557a79fab59b5cf7a12f11a9559540b9a1e223657f35f613",
   "groups": {
     "saudi_schoolbooks": {
       "documents": 229,

results/arabic_4bench_by_benchmark.csv

@@ -1,6 +1,6 @@
 benchmark,tokenizer,documents,tokens,tokens_per_word,characters_per_token,bytes_per_token,token_ratio_vs_yours,extra_tokens_vs_yours_pct
 Arabic-Dialects,ALLaM-7B,9807,432591,1.2811703139021422,4.223703220825214,7.658263810389028,0.9822639315897748,-1.7736068410225236
-Arabic-Dialects,SunAI-64K,9807,440402,1.3043035305476331,4.148791331556169,7.522436319544417,1.0,0.0
+Arabic-Dialects,SamAI-Tokenizer,9807,440402,1.3043035305476331,4.148791331556169,7.522436319544417,1.0,0.0
 Arabic-Dialects,Falcon-H1-7B,9807,545003,1.6140919820051947,3.352524665001844,6.078674796285525,1.237512545356288,23.751254535628807
 Arabic-Dialects,Mistral-Small-3.1,9807,566146,1.676709521313301,3.2273229873566183,5.851663705122,1.2855209558539697,28.552095585396977
 Arabic-Dialects,Qwen3.8-Flash-Next,9807,571810,1.6934841390421527,3.195355100470436,5.7937007047795595,1.298381932870423,29.83819328704229
@@ -18,7 +18,7 @@ ArabicMMLU,ALLaM-7B,14435,818785,1.9942835013128217,3.17674725355252,5.800798744
 ArabicMMLU,Qwen3.5-4B,14435,984657,2.3982916266812158,2.641603116618274,4.823615736241148,0.9655753009035442,-3.4424699096455846
 ArabicMMLU,Qwen3.8-Flash-Next,14435,984657,2.3982916266812158,2.641603116618274,4.823615736241148,0.9655753009035442,-3.4424699096455846
 ArabicMMLU,Mistral-Small-3.1,14435,1008568,2.4565307404899577,2.578976330797725,4.709258076797995,0.9890229288794836,-1.097707112051638
-ArabicMMLU,SunAI-64K,14435,1019762,2.483795540790031,2.55066672419643,4.6575642159641175,1.0,0.0
+ArabicMMLU,SamAI-Tokenizer,14435,1019762,2.483795540790031,2.55066672419643,4.6575642159641175,1.0,0.0
 ArabicMMLU,Falcon-H1-7B,14435,1053222,2.5652927909276464,2.4696341322152406,4.509597216921029,1.0328115776034015,3.2811577603401476
 ArabicMMLU,GPT-OSS-20B,14435,1088017,2.650041649820005,2.390654741607898,4.365379401240974,1.0669322841996465,6.6932284199646475
 ArabicMMLU,tiktoken-o200k,14435,1088017,2.650041649820005,2.390654741607898,4.365379401240974,1.0669322841996465,6.6932284199646475
@@ -30,7 +30,7 @@ ArabicMMLU,GLM-5.3-Flash,14435,1368588,3.333417769615604,1.900552248010358,3.470
 ArabicMMLU,Kimi-K2.6,14435,1471044,3.5829659543167236,1.768181645144537,3.228732111344052,1.4425365918714366,44.253659187143654
 ArabicMMLU,tiktoken-cl100k,14435,1960187,4.774352966392736,1.3269514592230232,2.423037699974543,1.9222004742283003,92.22004742283004
 MASSIVE-Arabic,ALLaM-7B,2907,18333,1.1938655899973951,4.306005563737522,7.766813942071674,0.9513258263712314,-4.867417362876858
-MASSIVE-Arabic,SunAI-64K,2907,19271,1.2549492055222715,4.096414301281719,7.388770691712937,1.0,0.0
+MASSIVE-Arabic,SamAI-Tokenizer,2907,19271,1.2549492055222715,4.096414301281719,7.388770691712937,1.0,0.0
 MASSIVE-Arabic,Falcon-H1-7B,2907,23860,1.5537900494920551,3.308549874266555,5.967686504610226,1.2381298323906387,23.81298323906387
 MASSIVE-Arabic,Mistral-Small-3.1,2907,24875,1.619887991664496,3.1735477386934674,5.724180904522613,1.2907996471381868,29.079964713818683
 MASSIVE-Arabic,Qwen3.8-Flash-Next,2907,24973,1.6262698619432143,3.16109398150002,5.7017178552837064,1.295885008562088,29.58850085620881
@@ -44,7 +44,7 @@ MASSIVE-Arabic,DeepSeek-V3,2907,30202,1.966788226100547,2.6138004105688366,4.714
 MASSIVE-Arabic,GLM-5.3-Flash,2907,32085,2.089411305027351,2.4604020570359983,4.437868162692848,1.6649369518966322,66.49369518966321
 MASSIVE-Arabic,Kimi-K2.6,2907,37595,2.448228705392029,2.0998005053863547,3.7874451389812474,1.950858803383322,95.0858803383322
 MASSIVE-Arabic,tiktoken-cl100k,2907,55108,3.5886949726491273,1.4324961893010089,2.583817231617914,2.859633646411707,185.9633646411707
-MLQA-Arabic,SunAI-64K,10933,919527,1.4396672657928193,4.123509151987924,7.390784609913575,1.0,0.0
+MLQA-Arabic,SamAI-Tokenizer,10933,919527,1.4396672657928193,4.123509151987924,7.390784609913575,1.0,0.0
 MLQA-Arabic,ALLaM-7B,10933,982992,1.5390319206898928,3.8572826635415143,6.913612725230724,1.0690191805134597,6.901918051345968
 MLQA-Arabic,Mistral-Small-3.1,10933,1161531,1.8185634123887597,3.2643795129015065,5.850920896644171,1.2631831365473771,26.318313654737715
 MLQA-Arabic,Falcon-H1-7B,10933,1161685,1.8188045241330937,3.263946766980722,5.850145263130711,1.2633506139569584,26.335061395695835

results/arabic_4bench_by_category.csv

@@ -1,5 +1,5 @@
 benchmark,group,subject,tokenizer,documents,words,tokens,tokens_per_word,bytes_per_token,token_ratio_vs_yours
-Arabic-Dialects,dialectal-Arabic,Egypt,SunAI-64K,3823,106839,159774,1.4954651391345857,6.495474858237261,1.0
+Arabic-Dialects,dialectal-Arabic,Egypt,SamAI-Tokenizer,3823,106839,159774,1.4954651391345857,6.495474858237261,1.0
 Arabic-Dialects,dialectal-Arabic,Egypt,Fanar-1,3823,106839,214173,2.0046331395838597,4.8456528133798376,1.3404746704720418
 Arabic-Dialects,dialectal-Arabic,Egypt,ALLaM-7B,3823,106839,149225,1.396727786669662,6.954652370581337,0.933975490380162
 Arabic-Dialects,dialectal-Arabic,Egypt,Qwen2.5-7B,3823,106839,241900,2.2641544754256406,4.2902356345597354,1.5140135441310851
@@ -14,7 +14,7 @@ Arabic-Dialects,dialectal-Arabic,Egypt,Mistral-Small-3.1,3823,106839,205238,1.92
 Arabic-Dialects,dialectal-Arabic,Egypt,Falcon-H1-7B,3823,106839,198397,1.8569717050889656,5.230966194045273,1.241735200971372
 Arabic-Dialects,dialectal-Arabic,Egypt,tiktoken-cl100k,3823,106839,410830,3.845318657044712,2.526125161258915,2.571319488777899
 Arabic-Dialects,dialectal-Arabic,Egypt,tiktoken-o200k,3823,106839,205328,1.9218450191409504,5.054391023143459,1.2851152252556737
-Arabic-Dialects,dialectal-Arabic,GULF,SunAI-64K,1650,64055,75770,1.182889704160487,8.243803616206941,1.0
+Arabic-Dialects,dialectal-Arabic,GULF,SamAI-Tokenizer,1650,64055,75770,1.182889704160487,8.243803616206941,1.0
 Arabic-Dialects,dialectal-Arabic,GULF,Fanar-1,1650,64055,113787,1.7763952853016938,5.489493527380105,1.5017421142932559
 Arabic-Dialects,dialectal-Arabic,GULF,ALLaM-7B,1650,64055,76783,1.1987042385450004,8.135042913144837,1.0133694074171835
 Arabic-Dialects,dialectal-Arabic,GULF,Qwen2.5-7B,1650,64055,122714,1.9157598938412301,5.090152712811904,1.6195591922924641
@@ -29,7 +29,7 @@ Arabic-Dialects,dialectal-Arabic,GULF,Mistral-Small-3.1,1650,64055,97667,1.52473
 Arabic-Dialects,dialectal-Arabic,GULF,Falcon-H1-7B,1650,64055,93907,1.4660369994535947,6.651612765821504,1.2393691434604726
 Arabic-Dialects,dialectal-Arabic,GULF,tiktoken-cl100k,1650,64055,242159,3.78048552025603,2.579433347511346,3.1959746601557346
 Arabic-Dialects,dialectal-Arabic,GULF,tiktoken-o200k,1650,64055,102956,1.6073062212161424,6.066989782042814,1.358796357397387
-Arabic-Dialects,dialectal-Arabic,Levant,SunAI-64K,1723,66179,80874,1.2220492905604496,7.875139105274872,1.0
+Arabic-Dialects,dialectal-Arabic,Levant,SamAI-Tokenizer,1723,66179,80874,1.2220492905604496,7.875139105274872,1.0
 Arabic-Dialects,dialectal-Arabic,Levant,Fanar-1,1723,66179,118308,1.7876970035811965,5.383355309869155,1.4628681652941613
 Arabic-Dialects,dialectal-Arabic,Levant,ALLaM-7B,1723,66179,81003,1.2239985493887788,7.862597681567349,1.0015950738185326
 Arabic-Dialects,dialectal-Arabic,Levant,Qwen2.5-7B,1723,66179,128485,1.9414769035494643,4.956952173405456,1.588705888171724
@@ -44,7 +44,7 @@ Arabic-Dialects,dialectal-Arabic,Levant,Mistral-Small-3.1,1723,66179,102815,1.55
 Arabic-Dialects,dialectal-Arabic,Levant,Falcon-H1-7B,1723,66179,99906,1.5096329651399991,6.374932436490301,1.2353290303434974
 Arabic-Dialects,dialectal-Arabic,Levant,tiktoken-cl100k,1723,66179,248795,3.7594251953036464,2.559914789284351,3.076328609936444
 Arabic-Dialects,dialectal-Arabic,Levant,tiktoken-o200k,1723,66179,108465,1.6389640218196104,5.871884939842346,1.3411603234661325
-Arabic-Dialects,dialectal-Arabic,MSA,SunAI-64K,999,48987,56035,1.1438749055871966,9.108557151780138,1.0
+Arabic-Dialects,dialectal-Arabic,MSA,SamAI-Tokenizer,999,48987,56035,1.1438749055871966,9.108557151780138,1.0
 Arabic-Dialects,dialectal-Arabic,MSA,Fanar-1,999,48987,87850,1.7933329250617511,5.809880478087649,1.5677701436602123
 Arabic-Dialects,dialectal-Arabic,MSA,ALLaM-7B,999,48987,57337,1.1704533855920958,8.901721401538273,1.0232354778263586
 Arabic-Dialects,dialectal-Arabic,MSA,Qwen2.5-7B,999,48987,94511,1.929307775532284,5.400408418067738,1.686642277148211
@@ -59,7 +59,7 @@ Arabic-Dialects,dialectal-Arabic,MSA,Mistral-Small-3.1,999,48987,73514,1.5006838
 Arabic-Dialects,dialectal-Arabic,MSA,Falcon-H1-7B,999,48987,68475,1.3978198297507503,7.453786053304126,1.2220041045774963
 Arabic-Dialects,dialectal-Arabic,MSA,tiktoken-cl100k,999,48987,193992,3.9600710392553125,2.6310260216916164,3.4619791201927366
 Arabic-Dialects,dialectal-Arabic,MSA,tiktoken-o200k,999,48987,78886,1.6103456018943803,6.470070734984661,1.4077986972427947
-Arabic-Dialects,dialectal-Arabic,North Africa,SunAI-64K,1612,51593,67949,1.3170197507413797,7.405009639582628,1.0
+Arabic-Dialects,dialectal-Arabic,North Africa,SamAI-Tokenizer,1612,51593,67949,1.3170197507413797,7.405009639582628,1.0
 Arabic-Dialects,dialectal-Arabic,North Africa,Fanar-1,1612,51593,98364,1.9065377086038804,5.115316579236306,1.4476151231070362
 Arabic-Dialects,dialectal-Arabic,North Africa,ALLaM-7B,1612,51593,68243,1.3227181982051828,7.373107864542884,1.004326774492634
 Arabic-Dialects,dialectal-Arabic,North Africa,Qwen2.5-7B,1612,51593,107869,2.0907681274591514,4.664574622922248,1.5874994481154985
@@ -74,7 +74,7 @@ Arabic-Dialects,dialectal-Arabic,North Africa,Mistral-Small-3.1,1612,51593,86912
 Arabic-Dialects,dialectal-Arabic,North Africa,Falcon-H1-7B,1612,51593,84318,1.6342914736495262,5.967444673735145,1.2409012641834316
 Arabic-Dialects,dialectal-Arabic,North Africa,tiktoken-cl100k,1612,51593,198645,3.8502316205686817,2.5329759118024615,2.9234425819364525
 Arabic-Dialects,dialectal-Arabic,North Africa,tiktoken-o200k,1612,51593,91599,1.775415269513306,5.493105819932532,1.348055159016321
-ArabicMMLU,Humanities,History,SunAI-64K,1065,22191,33607,1.5144427921229326,7.207873359716726,1.0
+ArabicMMLU,Humanities,History,SamAI-Tokenizer,1065,22191,33607,1.5144427921229326,7.207873359716726,1.0
 ArabicMMLU,Humanities,History,Fanar-1,1065,22191,48439,2.1828218647199313,5.000825780879044,1.4413366262980927
 ArabicMMLU,Humanities,History,ALLaM-7B,1065,22191,36070,1.6254337343968275,6.715691710562795,1.0732883030321063
 ArabicMMLU,Humanities,History,Qwen2.5-7B,1065,22191,51053,2.3006173674011987,4.744775037705914,1.5191180408843397
@@ -89,7 +89,7 @@ ArabicMMLU,Humanities,History,Mistral-Small-3.1,1065,22191,42584,1.9189761615069
 ArabicMMLU,Humanities,History,Falcon-H1-7B,1065,22191,42137,1.8988328601685367,5.748748131096186,1.2538161692504537
 ArabicMMLU,Humanities,History,tiktoken-cl100k,1065,22191,92850,4.184128700824659,2.608885298869144,2.762817270211563
 ArabicMMLU,Humanities,History,tiktoken-o200k,1065,22191,43976,1.98170429453382,5.5083454611606335,1.308536911952867
-ArabicMMLU,Humanities,Islamic Studies,SunAI-64K,2207,39137,61509,1.5716329815775354,6.14719797102863,1.0
+ArabicMMLU,Humanities,Islamic Studies,SamAI-Tokenizer,2207,39137,61509,1.5716329815775354,6.14719797102863,1.0
 ArabicMMLU,Humanities,Islamic Studies,Fanar-1,2207,39137,81788,2.0897871579323914,4.623025382696728,1.32969158984864
 ArabicMMLU,Humanities,Islamic Studies,ALLaM-7B,2207,39137,62283,1.5914096634897923,6.070805837869082,1.0125835243622885
 ArabicMMLU,Humanities,Islamic Studies,Qwen2.5-7B,2207,39137,85915,2.1952372435291414,4.400954431705756,1.3967874619974312
@@ -104,7 +104,7 @@ ArabicMMLU,Humanities,Islamic Studies,Mistral-Small-3.1,2207,39137,73091,1.86756
 ArabicMMLU,Humanities,Islamic Studies,Falcon-H1-7B,2207,39137,75517,1.9295551524133174,5.006925592912854,1.2277390300606414
 ArabicMMLU,Humanities,Islamic Studies,tiktoken-cl100k,2207,39137,152917,3.9072233436390116,2.4726354819934997,2.486091466289486
 ArabicMMLU,Humanities,Islamic Studies,tiktoken-o200k,2207,39137,76151,1.945754656718706,4.965240115034602,1.2380464647449967
-ArabicMMLU,Humanities,Law,SunAI-64K,314,11281,15675,1.389504476553497,7.622838915470495,1.0
+ArabicMMLU,Humanities,Law,SamAI-Tokenizer,314,11281,15675,1.389504476553497,7.622838915470495,1.0
 ArabicMMLU,Humanities,Law,Fanar-1,314,11281,23510,2.0840351032709865,5.082433007230965,1.4998405103668262
 ArabicMMLU,Humanities,Law,ALLaM-7B,314,11281,15815,1.4019147238719971,7.5553588365475814,1.0089314194577352
 ArabicMMLU,Humanities,Law,Qwen2.5-7B,314,11281,24968,2.213278964630795,4.785645626401794,1.5928548644338119
@@ -119,7 +119,7 @@ ArabicMMLU,Humanities,Law,Mistral-Small-3.1,314,11281,20244,1.7945217622551193,5
 ArabicMMLU,Humanities,Law,Falcon-H1-7B,314,11281,19290,1.709954791241911,6.194297563504406,1.230622009569378
 ArabicMMLU,Humanities,Law,tiktoken-cl100k,314,11281,47266,4.189876783973052,2.5279905217280922,3.0153748006379586
 ArabicMMLU,Humanities,Law,tiktoken-o200k,314,11281,21664,1.920397127914192,5.515509601181684,1.38207336523126
-ArabicMMLU,Humanities,Philosophy,SunAI-64K,39,1407,2160,1.535181236673774,6.564814814814815,1.0
+ArabicMMLU,Humanities,Philosophy,SamAI-Tokenizer,39,1407,2160,1.535181236673774,6.564814814814815,1.0
 ArabicMMLU,Humanities,Philosophy,Fanar-1,39,1407,2989,2.1243781094527363,4.744061559049849,1.3837962962962962
 ArabicMMLU,Humanities,Philosophy,ALLaM-7B,39,1407,2229,1.584221748400853,6.36159712875729,1.0319444444444446
 ArabicMMLU,Humanities,Philosophy,Qwen2.5-7B,39,1407,3076,2.1862117981520965,4.609882964889467,1.424074074074074
@@ -134,7 +134,7 @@ ArabicMMLU,Humanities,Philosophy,Mistral-Small-3.1,39,1407,2573,1.82871357498223
 ArabicMMLU,Humanities,Philosophy,Falcon-H1-7B,39,1407,2839,2.017768301350391,4.9947164494540335,1.3143518518518518
 ArabicMMLU,Humanities,Philosophy,tiktoken-cl100k,39,1407,5530,3.9303482587064678,2.5641952983725136,2.560185185185185
 ArabicMMLU,Humanities,Philosophy,tiktoken-o200k,39,1407,2701,1.919687277896233,5.249907441688263,1.2504629629629629
-ArabicMMLU,Language,Arabic Language,SunAI-64K,669,17268,32934,1.9072272411396802,5.384982085382887,1.0
+ArabicMMLU,Language,Arabic Language,SamAI-Tokenizer,669,17268,32934,1.9072272411396802,5.384982085382887,1.0
 ArabicMMLU,Language,Arabic Language,Fanar-1,669,17268,41489,2.402652304841325,4.274602906794572,1.2597619481387017
 ArabicMMLU,Language,Arabic Language,ALLaM-7B,669,17268,31908,1.847810979847116,5.558135890685722,0.968846784478047
 ArabicMMLU,Language,Arabic Language,Qwen2.5-7B,669,17268,43938,2.5444753300903407,4.036346670308162,1.3341227910366187
@@ -149,7 +149,7 @@ ArabicMMLU,Language,Arabic Language,Mistral-Small-3.1,669,17268,36795,2.13082001
 ArabicMMLU,Language,Arabic Language,Falcon-H1-7B,669,17268,38668,2.239286541579801,4.586453915382228,1.1741057873322402
 ArabicMMLU,Language,Arabic Language,tiktoken-cl100k,669,17268,72894,4.221334259902711,2.43297116360743,2.213335762433959
 ArabicMMLU,Language,Arabic Language,tiktoken-o200k,669,17268,38810,2.2475098447996293,4.569672764751353,1.178417440942491
-ArabicMMLU,Language,Arabic Language (General),SunAI-64K,612,130856,563958,4.309760347251941,3.265697090918118,1.0
+ArabicMMLU,Language,Arabic Language (General),SamAI-Tokenizer,612,130856,563958,4.309760347251941,3.265697090918118,1.0
 ArabicMMLU,Language,Arabic Language (General),Fanar-1,612,130856,544644,4.162162988323042,3.3815042486468223,0.9657527688232103
 ArabicMMLU,Language,Arabic Language (General),ALLaM-7B,612,130856,340840,2.6046952375129915,5.403462034972421,0.6043712475042468
 ArabicMMLU,Language,Arabic Language (General),Qwen2.5-7B,612,130856,638412,4.878736932200281,2.8848392574074424,1.132020469609439
@@ -164,7 +164,7 @@ ArabicMMLU,Language,Arabic Language (General),Mistral-Small-3.1,612,130856,44657
 ArabicMMLU,Language,Arabic Language (General),Falcon-H1-7B,612,130856,485719,3.7118588371950847,3.7917314331949132,0.86126803769075
 ArabicMMLU,Language,Arabic Language (General),tiktoken-cl100k,612,130856,800767,6.119451916610625,2.2999399325896297,1.419905383024977
 ArabicMMLU,Language,Arabic Language (General),tiktoken-o200k,612,130856,507338,3.877070978785841,3.6301558329949657,0.8996024526649148
-ArabicMMLU,Language,Arabic Language (Grammar),SunAI-64K,365,9238,18334,1.9846287075124487,5.132158830587978,1.0
+ArabicMMLU,Language,Arabic Language (Grammar),SamAI-Tokenizer,365,9238,18334,1.9846287075124487,5.132158830587978,1.0
 ArabicMMLU,Language,Arabic Language (Grammar),Fanar-1,365,9238,23311,2.523381684347261,4.03642057397795,1.2714628558961492
 ArabicMMLU,Language,Arabic Language (Grammar),ALLaM-7B,365,9238,16880,1.8272353323230137,5.574229857819906,0.9206937929529835
 ArabicMMLU,Language,Arabic Language (Grammar),Qwen2.5-7B,365,9238,24349,2.6357436674604893,3.8643476118115734,1.3280789789462202
@@ -179,7 +179,7 @@ ArabicMMLU,Language,Arabic Language (Grammar),Mistral-Small-3.1,365,9238,19726,2
 ArabicMMLU,Language,Arabic Language (Grammar),Falcon-H1-7B,365,9238,21652,2.343797358735657,4.345695547755404,1.1809752372640996
 ArabicMMLU,Language,Arabic Language (Grammar),tiktoken-cl100k,365,9238,39172,4.240311755791296,2.402047380782191,2.136576851750845
 ArabicMMLU,Language,Arabic Language (Grammar),tiktoken-o200k,365,9238,21320,2.307858843905607,4.413367729831145,1.1628668048434603
-ArabicMMLU,Other,Driving Test,SunAI-64K,1211,29647,44828,1.5120585556717374,6.785602748282323,1.0
+ArabicMMLU,Other,Driving Test,SamAI-Tokenizer,1211,29647,44828,1.5120585556717374,6.785602748282323,1.0
 ArabicMMLU,Other,Driving Test,Fanar-1,1211,29647,63874,2.15448443350086,4.762266336850675,1.4248683858302846
 ArabicMMLU,Other,Driving Test,ALLaM-7B,1211,29647,47040,1.5866698148210612,6.466517857142857,1.0493441599000624
 ArabicMMLU,Other,Driving Test,Qwen2.5-7B,1211,29647,68843,2.3220899247815967,4.4185320221373265,1.5357142857142858
@@ -194,7 +194,7 @@ ArabicMMLU,Other,Driving Test,Mistral-Small-3.1,1211,29647,55563,1.8741525280804
 ArabicMMLU,Other,Driving Test,Falcon-H1-7B,1211,29647,54844,1.849900495834317,5.546367879804537,1.223431783706612
 ArabicMMLU,Other,Driving Test,tiktoken-cl100k,1211,29647,119858,4.042837386582116,2.5378781558177175,2.6737307040242704
 ArabicMMLU,Other,Driving Test,tiktoken-o200k,1211,29647,57716,1.9467737039160793,5.270375632406958,1.2874988846256803
-ArabicMMLU,Other,General Knowledge,SunAI-64K,1192,18878,32359,1.7141116643712258,6.052319292932414,1.0
+ArabicMMLU,Other,General Knowledge,SamAI-Tokenizer,1192,18878,32359,1.7141116643712258,6.052319292932414,1.0
 ArabicMMLU,Other,General Knowledge,Fanar-1,1192,18878,43988,2.330119716071618,4.452282440665636,1.3593745171358818
 ArabicMMLU,Other,General Knowledge,ALLaM-7B,1192,18878,33899,1.7956881025532365,5.777368063954689,1.0475910874872525
 ArabicMMLU,Other,General Knowledge,Qwen2.5-7B,1192,18878,46517,2.464085178514673,4.2102242191026935,1.4375289718470905
@@ -209,7 +209,7 @@ ArabicMMLU,Other,General Knowledge,Mistral-Small-3.1,1192,18878,39264,2.07988134
 ArabicMMLU,Other,General Knowledge,Falcon-H1-7B,1192,18878,39995,2.118603665642547,4.896787098387298,1.2359776260082203
 ArabicMMLU,Other,General Knowledge,tiktoken-cl100k,1192,18878,79929,4.233976056785677,2.4502621076205133,2.4700701504990885
 ArabicMMLU,Other,General Knowledge,tiktoken-o200k,1192,18878,40593,2.1502807500794576,4.8246495701229275,1.2544578015389845
-ArabicMMLU,Other,Management,SunAI-64K,75,1686,2805,1.6637010676156583,6.080213903743315,1.0
+ArabicMMLU,Other,Management,SamAI-Tokenizer,75,1686,2805,1.6637010676156583,6.080213903743315,1.0
 ArabicMMLU,Other,Management,Fanar-1,75,1686,3825,2.2686832740213525,4.458823529411765,1.3636363636363635
 ArabicMMLU,Other,Management,ALLaM-7B,75,1686,2906,1.723606168446026,5.868891947694426,1.036007130124777
 ArabicMMLU,Other,Management,Qwen2.5-7B,75,1686,3811,2.260379596678529,4.475203358698504,1.3586452762923351
@@ -224,7 +224,7 @@ ArabicMMLU,Other,Management,Mistral-Small-3.1,75,1686,3247,1.9258600237247925,5.
 ArabicMMLU,Other,Management,Falcon-H1-7B,75,1686,3347,1.9851720047449586,5.0956080071706005,1.1932263814616755
 ArabicMMLU,Other,Management,tiktoken-cl100k,75,1686,6753,4.005338078291815,2.525544202576633,2.407486631016043
 ArabicMMLU,Other,Management,tiktoken-o200k,75,1686,3395,2.013641755634638,5.023564064801178,1.2103386809269163
-ArabicMMLU,STEM,Biology,SunAI-64K,1403,28953,47209,1.6305391496563395,6.445698913342794,1.0
+ArabicMMLU,STEM,Biology,SamAI-Tokenizer,1403,28953,47209,1.6305391496563395,6.445698913342794,1.0
 ArabicMMLU,STEM,Biology,Fanar-1,1403,28953,67514,2.331848167720098,4.507139260005332,1.4301086657205193
 ArabicMMLU,STEM,Biology,ALLaM-7B,1403,28953,50960,1.7600939453597209,5.971251962323391,1.0794551886292867
 ArabicMMLU,STEM,Biology,Qwen2.5-7B,1403,28953,72055,2.4886885642247782,4.223093470265769,1.5262979516617594
@@ -239,7 +239,7 @@ ArabicMMLU,STEM,Biology,Mistral-Small-3.1,1403,28953,60317,2.0832728905467484,5.
 ArabicMMLU,STEM,Biology,Falcon-H1-7B,1403,28953,61273,2.1162919213898386,4.9662167675811535,1.2979092969560888
 ArabicMMLU,STEM,Biology,tiktoken-cl100k,1403,28953,120974,4.1782889510586125,2.5153752045894158,2.562519858501557
 ArabicMMLU,STEM,Biology,tiktoken-o200k,1403,28953,62898,2.172417366076054,4.837912175267894,1.3323306996547268
-ArabicMMLU,STEM,Computer Science,SunAI-64K,542,9776,14540,1.487315875613748,7.119463548830812,1.0
+ArabicMMLU,STEM,Computer Science,SamAI-Tokenizer,542,9776,14540,1.487315875613748,7.119463548830812,1.0
 ArabicMMLU,STEM,Computer Science,Fanar-1,542,9776,21367,2.1856587561374794,4.844713811016989,1.469532324621733
 ArabicMMLU,STEM,Computer Science,ALLaM-7B,542,9776,15718,1.6078150572831424,6.585888789922382,1.0810178817056395
 ArabicMMLU,STEM,Computer Science,Qwen2.5-7B,542,9776,22331,2.2842675941080195,4.635573865926291,1.5358321870701512
@@ -254,7 +254,7 @@ ArabicMMLU,STEM,Computer Science,Mistral-Small-3.1,542,9776,18788,1.921849427168
 ArabicMMLU,STEM,Computer Science,Falcon-H1-7B,542,9776,19322,1.9764729950900164,5.357468170996791,1.3288858321870702
 ArabicMMLU,STEM,Computer Science,tiktoken-cl100k,542,9776,40046,4.096358428805237,2.584952304849423,2.7541953232462175
 ArabicMMLU,STEM,Computer Science,tiktoken-o200k,542,9776,19491,1.9937602291325696,5.311015340413524,1.34050894085282
-ArabicMMLU,STEM,Math,SunAI-64K,409,4585,7744,1.6889858233369683,4.202221074380165,1.0
+ArabicMMLU,STEM,Math,SamAI-Tokenizer,409,4585,7744,1.6889858233369683,4.202221074380165,1.0
 ArabicMMLU,STEM,Math,Fanar-1,409,4585,11268,2.4575790621592146,2.8880014199503017,1.4550619834710743
 ArabicMMLU,STEM,Math,ALLaM-7B,409,4585,9977,2.176008724100327,3.2617019144031274,1.2883522727272727
 ArabicMMLU,STEM,Math,Qwen2.5-7B,409,4585,11670,2.5452562704471102,2.788517566409597,1.5069731404958677
@@ -269,7 +269,7 @@ ArabicMMLU,STEM,Math,Mistral-Small-3.1,409,4585,10655,2.3238822246455832,3.05415
 ArabicMMLU,STEM,Math,Falcon-H1-7B,409,4585,11069,2.414176663031625,2.939922305537989,1.4293646694214877
 ArabicMMLU,STEM,Math,tiktoken-cl100k,409,4585,14486,3.159432933478735,2.2464448432969766,1.8706095041322315
 ArabicMMLU,STEM,Math,tiktoken-o200k,409,4585,9200,2.0065430752453652,3.5371739130434783,1.18801652892562
-ArabicMMLU,STEM,Natural Science,SunAI-64K,578,9714,17306,1.7815523985999588,6.16751415693979,1.0
+ArabicMMLU,STEM,Natural Science,SamAI-Tokenizer,578,9714,17306,1.7815523985999588,6.16751415693979,1.0
 ArabicMMLU,STEM,Natural Science,Fanar-1,578,9714,23515,2.4207329627341982,4.539017648309589,1.3587773026695944
 ArabicMMLU,STEM,Natural Science,ALLaM-7B,578,9714,17754,1.8276714021000617,6.011884645713642,1.0258869756153934
 ArabicMMLU,STEM,Natural Science,Qwen2.5-7B,578,9714,25268,2.601194152769199,4.224117460820009,1.460071651450364
@@ -284,7 +284,7 @@ ArabicMMLU,STEM,Natural Science,Mistral-Small-3.1,578,9714,20500,2.1103561869466
 ArabicMMLU,STEM,Natural Science,Falcon-H1-7B,578,9714,21462,2.209388511426807,4.973208461466778,1.240147925574945
 ArabicMMLU,STEM,Natural Science,tiktoken-cl100k,578,9714,42341,4.358760551780935,2.5208426820339622,2.446608112793251
 ArabicMMLU,STEM,Natural Science,tiktoken-o200k,578,9714,21630,2.2266831377393452,4.934581599630143,1.2498555414307178
-ArabicMMLU,STEM,Physics,SunAI-64K,255,7398,11407,1.5419032170856988,5.738406241781362,1.0
+ArabicMMLU,STEM,Physics,SamAI-Tokenizer,255,7398,11407,1.5419032170856988,5.738406241781362,1.0
 ArabicMMLU,STEM,Physics,Fanar-1,255,7398,16104,2.1768045417680453,4.064704421261799,1.411764705882353
 ArabicMMLU,STEM,Physics,ALLaM-7B,255,7398,12596,1.7026223303595567,5.196729120355668,1.1042342421320241
 ArabicMMLU,STEM,Physics,Qwen2.5-7B,255,7398,17613,2.380778588807786,3.7164594333730765,1.5440518979573945
@@ -299,7 +299,7 @@ ArabicMMLU,STEM,Physics,Mistral-Small-3.1,255,7398,14316,1.935117599351176,4.572
 ArabicMMLU,STEM,Physics,Falcon-H1-7B,255,7398,14855,2.0079751284130847,4.406462470548637,1.302270535636013
 ArabicMMLU,STEM,Physics,tiktoken-cl100k,255,7398,27688,3.742633144092998,2.3641288644900316,2.4272814938195846
 ArabicMMLU,STEM,Physics,tiktoken-o200k,255,7398,14807,2.0014868883482024,4.420746944012967,1.2980625931445604
-ArabicMMLU,Social Science,Accounting,SunAI-64K,74,2202,4070,1.8483197093551318,5.774447174447174,1.0
+ArabicMMLU,Social Science,Accounting,SamAI-Tokenizer,74,2202,4070,1.8483197093551318,5.774447174447174,1.0
 ArabicMMLU,Social Science,Accounting,Fanar-1,74,2202,5447,2.4736603088101727,4.314668624931155,1.3383292383292382
 ArabicMMLU,Social Science,Accounting,ALLaM-7B,74,2202,4242,1.9264305177111716,5.5403111739745405,1.0422604422604422
 ArabicMMLU,Social Science,Accounting,Qwen2.5-7B,74,2202,5808,2.637602179836512,4.046487603305785,1.427027027027027
@@ -314,7 +314,7 @@ ArabicMMLU,Social Science,Accounting,Mistral-Small-3.1,74,2202,4959,2.2520435967
 ArabicMMLU,Social Science,Accounting,Falcon-H1-7B,74,2202,5037,2.287465940054496,4.665872543180464,1.2375921375921375
 ArabicMMLU,Social Science,Accounting,tiktoken-cl100k,74,2202,9647,4.381017257039056,2.436197781693791,2.3702702702702703
 ArabicMMLU,Social Science,Accounting,tiktoken-o200k,74,2202,5048,2.292461398728429,4.655705229793978,1.2402948402948404
-ArabicMMLU,Social Science,Civics,SunAI-64K,323,7045,10769,1.5286018452803407,7.218590398365679,1.0
+ArabicMMLU,Social Science,Civics,SamAI-Tokenizer,323,7045,10769,1.5286018452803407,7.218590398365679,1.0
 ArabicMMLU,Social Science,Civics,Fanar-1,323,7045,15673,2.2246983676366217,4.9599310916863395,1.455381186739716
 ArabicMMLU,Social Science,Civics,ALLaM-7B,323,7045,11696,1.6601845280340668,6.6464603283173735,1.0860804160089146
 ArabicMMLU,Social Science,Civics,Qwen2.5-7B,323,7045,16640,2.361958836053939,4.6716947115384615,1.5451759680564583
@@ -329,7 +329,7 @@ ArabicMMLU,Social Science,Civics,Mistral-Small-3.1,323,7045,13846,1.965365507452
 ArabicMMLU,Social Science,Civics,Falcon-H1-7B,323,7045,13354,1.895528743789922,5.8212520593080725,1.2400408580183861
 ArabicMMLU,Social Science,Civics,tiktoken-cl100k,323,7045,30289,4.299361249112846,2.5665092938030307,2.812610270220076
 ArabicMMLU,Social Science,Civics,tiktoken-o200k,323,7045,14008,1.988360539389638,5.549471730439748,1.3007707308013743
-ArabicMMLU,Social Science,Economics,SunAI-64K,584,15822,24265,1.5336240677537605,7.296723676076653,1.0
+ArabicMMLU,Social Science,Economics,SamAI-Tokenizer,584,15822,24265,1.5336240677537605,7.296723676076653,1.0
 ArabicMMLU,Social Science,Economics,Fanar-1,584,15822,35489,2.2430160535962584,4.989010679365437,1.462559241706161
 ArabicMMLU,Social Science,Economics,ALLaM-7B,584,15822,25510,1.6123119706737454,6.9406115248922,1.0513084689882546
 ArabicMMLU,Social Science,Economics,Qwen2.5-7B,584,15822,37455,2.3672734167614715,4.727139233747163,1.5435812899237584
@@ -344,7 +344,7 @@ ArabicMMLU,Social Science,Economics,Mistral-Small-3.1,584,15822,31126,1.96726077
 ArabicMMLU,Social Science,Economics,Falcon-H1-7B,584,15822,29581,1.8696119327518645,5.985429836719516,1.2190809808365959
 ArabicMMLU,Social Science,Economics,tiktoken-cl100k,584,15822,67099,4.2408671470104915,2.6387129465416774,2.7652586029260253
 ArabicMMLU,Social Science,Economics,tiktoken-o200k,584,15822,31515,1.9918467956010617,5.618118356338251,1.2987842571605193
-ArabicMMLU,Social Science,Geography,SunAI-64K,1362,27007,43819,1.6225052764098198,6.798717451333896,1.0
+ArabicMMLU,Social Science,Geography,SamAI-Tokenizer,1362,27007,43819,1.6225052764098198,6.798717451333896,1.0
 ArabicMMLU,Social Science,Geography,Fanar-1,1362,27007,62549,2.3160291776206168,4.762873906857024,1.4274401515324402
 ArabicMMLU,Social Science,Geography,ALLaM-7B,1362,27007,46757,1.7312918872884808,6.371516564364694,1.0670485405874164
 ArabicMMLU,Social Science,Geography,Qwen2.5-7B,1362,27007,67067,2.4833191394823566,4.442020665901263,1.530546110134873
@@ -359,7 +359,7 @@ ArabicMMLU,Social Science,Geography,Mistral-Small-3.1,1362,27007,54533,2.0192172
 ArabicMMLU,Social Science,Geography,Falcon-H1-7B,1362,27007,55042,2.0380642055763323,5.412466843501326,1.256121773659828
 ArabicMMLU,Social Science,Geography,tiktoken-cl100k,1362,27007,115722,4.284889102825193,2.574385164445827,2.6409091946415937
 ArabicMMLU,Social Science,Geography,tiktoken-o200k,1362,27007,56560,2.0942718554448847,5.2672029702970296,1.2907642803350146
-ArabicMMLU,Social Science,Political Science,SunAI-64K,210,3858,10023,2.5979782270606533,4.5571186271575375,1.0
+ArabicMMLU,Social Science,Political Science,SamAI-Tokenizer,210,3858,10023,2.5979782270606533,4.5571186271575375,1.0
 ArabicMMLU,Social Science,Political Science,Fanar-1,210,3858,12004,3.111456713322965,3.8050649783405532,1.1976454155442482
 ArabicMMLU,Social Science,Political Science,ALLaM-7B,210,3858,12164,3.152928978745464,3.7550147977638932,1.2136086999900229
 ArabicMMLU,Social Science,Political Science,Qwen2.5-7B,210,3858,12639,3.276049766718507,3.6138935042329297,1.2609997006884166
@@ -374,7 +374,7 @@ ArabicMMLU,Social Science,Political Science,Mistral-Small-3.1,210,3858,14988,3.8
 ArabicMMLU,Social Science,Political Science,Falcon-H1-7B,210,3858,12929,3.351218247796786,3.5328331657514114,1.2899331537463834
 ArabicMMLU,Social Science,Political Science,tiktoken-cl100k,210,3858,22315,5.784085018144116,2.046874299798342,2.226379327546643
 ArabicMMLU,Social Science,Political Science,tiktoken-o200k,210,3858,13506,3.5007776049766717,3.3819043388123795,1.3475007482789585
-ArabicMMLU,Social Science,Social Science,SunAI-64K,946,12617,20441,1.6201157168899105,6.404823638765227,1.0
+ArabicMMLU,Social Science,Social Science,SamAI-Tokenizer,946,12617,20441,1.6201157168899105,6.404823638765227,1.0
 ArabicMMLU,Social Science,Social Science,Fanar-1,946,12617,28272,2.2407862407862407,4.6307654216185625,1.383102587936011
 ArabicMMLU,Social Science,Social Science,ALLaM-7B,946,12617,21541,1.7072996750416105,6.0777586927255,1.0538134142165256
 ArabicMMLU,Social Science,Social Science,Qwen2.5-7B,946,12617,29450,2.3341523341523343,4.44553480475382,1.4407318624333447
@@ -389,7 +389,7 @@ ArabicMMLU,Social Science,Social Science,Mistral-Small-3.1,946,12617,24880,1.971
 ArabicMMLU,Social Science,Social Science,Falcon-H1-7B,946,12617,25290,2.0044384560513593,5.176789244760775,1.2372193141235752
 ArabicMMLU,Social Science,Social Science,tiktoken-cl100k,946,12617,51644,4.0932075770785445,2.535066997134227,2.526490876180226
 ArabicMMLU,Social Science,Social Science,tiktoken-o200k,946,12617,25690,2.036141713561068,5.096185286103542,1.2567878283841298
-MASSIVE-Arabic,spoken-NLU,alarm_query,SunAI-64K,34,208,292,1.4038461538461537,6.7465753424657535,1.0
+MASSIVE-Arabic,spoken-NLU,alarm_query,SamAI-Tokenizer,34,208,292,1.4038461538461537,6.7465753424657535,1.0
 MASSIVE-Arabic,spoken-NLU,alarm_query,Fanar-1,34,208,383,1.8413461538461537,5.143603133159269,1.3116438356164384
 MASSIVE-Arabic,spoken-NLU,alarm_query,ALLaM-7B,34,208,274,1.3173076923076923,7.189781021897811,0.9383561643835616
 MASSIVE-Arabic,spoken-NLU,alarm_query,Qwen2.5-7B,34,208,411,1.9759615384615385,4.793187347931873,1.4075342465753424
@@ -404,7 +404,7 @@ MASSIVE-Arabic,spoken-NLU,alarm_query,Mistral-Small-3.1,34,208,372,1.78846153846
 MASSIVE-Arabic,spoken-NLU,alarm_query,Falcon-H1-7B,34,208,361,1.7355769230769231,5.457063711911357,1.2363013698630136
 MASSIVE-Arabic,spoken-NLU,alarm_query,tiktoken-cl100k,34,208,751,3.610576923076923,2.623169107856192,2.5719178082191783
 MASSIVE-Arabic,spoken-NLU,alarm_query,tiktoken-o200k,34,208,366,1.7596153846153846,5.382513661202186,1.2534246575342465
-MASSIVE-Arabic,spoken-NLU,alarm_remove,SunAI-64K,19,77,120,1.5584415584415585,6.1,1.0
+MASSIVE-Arabic,spoken-NLU,alarm_remove,SamAI-Tokenizer,19,77,120,1.5584415584415585,6.1,1.0
 MASSIVE-Arabic,spoken-NLU,alarm_remove,Fanar-1,19,77,146,1.896103896103896,5.013698630136986,1.2166666666666666
 MASSIVE-Arabic,spoken-NLU,alarm_remove,ALLaM-7B,19,77,119,1.5454545454545454,6.151260504201681,0.9916666666666667
 MASSIVE-Arabic,spoken-NLU,alarm_remove,Qwen2.5-7B,19,77,158,2.051948051948052,4.632911392405063,1.3166666666666667
@@ -419,7 +419,7 @@ MASSIVE-Arabic,spoken-NLU,alarm_remove,Mistral-Small-3.1,19,77,144,1.87012987012
 MASSIVE-Arabic,spoken-NLU,alarm_remove,Falcon-H1-7B,19,77,142,1.844155844155844,5.154929577464789,1.1833333333333333
 MASSIVE-Arabic,spoken-NLU,alarm_remove,tiktoken-cl100k,19,77,283,3.675324675324675,2.5865724381625443,2.3583333333333334
 MASSIVE-Arabic,spoken-NLU,alarm_remove,tiktoken-o200k,19,77,138,1.7922077922077921,5.304347826086956,1.15
-MASSIVE-Arabic,spoken-NLU,alarm_set,SunAI-64K,41,229,313,1.3668122270742358,6.616613418530352,1.0
+MASSIVE-Arabic,spoken-NLU,alarm_set,SamAI-Tokenizer,41,229,313,1.3668122270742358,6.616613418530352,1.0
 MASSIVE-Arabic,spoken-NLU,alarm_set,Fanar-1,41,229,429,1.8733624454148472,4.827505827505828,1.3706070287539935
 MASSIVE-Arabic,spoken-NLU,alarm_set,ALLaM-7B,41,229,299,1.3056768558951966,6.926421404682274,0.9552715654952076
 MASSIVE-Arabic,spoken-NLU,alarm_set,Qwen2.5-7B,41,229,439,1.9170305676855894,4.71753986332574,1.402555910543131
@@ -434,7 +434,7 @@ MASSIVE-Arabic,spoken-NLU,alarm_set,Mistral-Small-3.1,41,229,386,1.6855895196506
 MASSIVE-Arabic,spoken-NLU,alarm_set,Falcon-H1-7B,41,229,365,1.5938864628820961,5.673972602739726,1.1661341853035143
 MASSIVE-Arabic,spoken-NLU,alarm_set,tiktoken-cl100k,41,229,821,3.5851528384279474,2.5225334957369063,2.623003194888179
 MASSIVE-Arabic,spoken-NLU,alarm_set,tiktoken-o200k,41,229,413,1.8034934497816595,5.014527845036319,1.3194888178913737
-MASSIVE-Arabic,spoken-NLU,audio_volume_down,SunAI-64K,9,36,45,1.25,7.088888888888889,1.0
+MASSIVE-Arabic,spoken-NLU,audio_volume_down,SamAI-Tokenizer,9,36,45,1.25,7.088888888888889,1.0
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,Fanar-1,9,36,60,1.6666666666666667,5.316666666666666,1.3333333333333333
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,ALLaM-7B,9,36,45,1.25,7.088888888888889,1.0
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,Qwen2.5-7B,9,36,72,2.0,4.430555555555555,1.6
@@ -449,7 +449,7 @@ MASSIVE-Arabic,spoken-NLU,audio_volume_down,Mistral-Small-3.1,9,36,55,1.52777777
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,Falcon-H1-7B,9,36,56,1.5555555555555556,5.696428571428571,1.2444444444444445
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,tiktoken-cl100k,9,36,130,3.611111111111111,2.453846153846154,2.888888888888889
 MASSIVE-Arabic,spoken-NLU,audio_volume_down,tiktoken-o200k,9,36,59,1.6388888888888888,5.406779661016949,1.3111111111111111
-MASSIVE-Arabic,spoken-NLU,audio_volume_mute,SunAI-64K,31,112,147,1.3125,6.809523809523809,1.0
+MASSIVE-Arabic,spoken-NLU,audio_volume_mute,SamAI-Tokenizer,31,112,147,1.3125,6.809523809523809,1.0
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,Fanar-1,31,112,198,1.7678571428571428,5.055555555555555,1.346938775510204
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,ALLaM-7B,31,112,134,1.1964285714285714,7.470149253731344,0.9115646258503401
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,Qwen2.5-7B,31,112,213,1.9017857142857142,4.699530516431925,1.4489795918367347
@@ -464,7 +464,7 @@ MASSIVE-Arabic,spoken-NLU,audio_volume_mute,Mistral-Small-3.1,31,112,185,1.65178
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,Falcon-H1-7B,31,112,182,1.625,5.5,1.2380952380952381
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,tiktoken-cl100k,31,112,406,3.625,2.4655172413793105,2.761904761904762
 MASSIVE-Arabic,spoken-NLU,audio_volume_mute,tiktoken-o200k,31,112,188,1.6785714285714286,5.324468085106383,1.2789115646258504
-MASSIVE-Arabic,spoken-NLU,audio_volume_other,SunAI-64K,6,26,33,1.2692307692307692,7.878787878787879,1.0
+MASSIVE-Arabic,spoken-NLU,audio_volume_other,SamAI-Tokenizer,6,26,33,1.2692307692307692,7.878787878787879,1.0
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,Fanar-1,6,26,44,1.6923076923076923,5.909090909090909,1.3333333333333333
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,ALLaM-7B,6,26,32,1.2307692307692308,8.125,0.9696969696969697
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,Qwen2.5-7B,6,26,53,2.0384615384615383,4.90566037735849,1.606060606060606
@@ -479,7 +479,7 @@ MASSIVE-Arabic,spoken-NLU,audio_volume_other,Mistral-Small-3.1,6,26,40,1.5384615
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,Falcon-H1-7B,6,26,40,1.5384615384615385,6.5,1.2121212121212122
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,tiktoken-cl100k,6,26,98,3.769230769230769,2.6530612244897958,2.9696969696969697
 MASSIVE-Arabic,spoken-NLU,audio_volume_other,tiktoken-o200k,6,26,42,1.6153846153846154,6.190476190476191,1.2727272727272727
-MASSIVE-Arabic,spoken-NLU,audio_volume_up,SunAI-64K,13,41,57,1.3902439024390243,6.631578947368421,1.0
+MASSIVE-Arabic,spoken-NLU,audio_volume_up,SamAI-Tokenizer,13,41,57,1.3902439024390243,6.631578947368421,1.0
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,Fanar-1,13,41,80,1.951219512195122,4.725,1.4035087719298245
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,ALLaM-7B,13,41,47,1.146341463414634,8.042553191489361,0.8245614035087719
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,Qwen2.5-7B,13,41,87,2.1219512195121952,4.344827586206897,1.5263157894736843
@@ -494,7 +494,7 @@ MASSIVE-Arabic,spoken-NLU,audio_volume_up,Mistral-Small-3.1,13,41,68,1.658536585
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,Falcon-H1-7B,13,41,71,1.7317073170731707,5.323943661971831,1.2456140350877194
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,tiktoken-cl100k,13,41,153,3.731707317073171,2.4705882352941178,2.6842105263157894
 MASSIVE-Arabic,spoken-NLU,audio_volume_up,tiktoken-o200k,13,41,75,1.829268292682927,5.04,1.3157894736842106
-MASSIVE-Arabic,spoken-NLU,calendar_query,SunAI-64K,125,659,797,1.2094081942336874,7.8833124215809285,1.0
+MASSIVE-Arabic,spoken-NLU,calendar_query,SamAI-Tokenizer,125,659,797,1.2094081942336874,7.8833124215809285,1.0
 MASSIVE-Arabic,spoken-NLU,calendar_query,Fanar-1,125,659,1123,1.7040971168437027,5.594835262689226,1.409033877038896
 MASSIVE-Arabic,spoken-NLU,calendar_query,ALLaM-7B,125,659,760,1.1532625189681336,8.267105263157895,0.9535759096612296
 MASSIVE-Arabic,spoken-NLU,calendar_query,Qwen2.5-7B,125,659,1216,1.8452200303490136,5.166940789473684,1.5257214554579674
@@ -509,7 +509,7 @@ MASSIVE-Arabic,spoken-NLU,calendar_query,Mistral-Small-3.1,125,659,992,1.5053110
 MASSIVE-Arabic,spoken-NLU,calendar_query,Falcon-H1-7B,125,659,967,1.4673748103186646,6.49741468459152,1.2132998745294856
 MASSIVE-Arabic,spoken-NLU,calendar_query,tiktoken-cl100k,125,659,2453,3.7223065250379364,2.5613534447615165,3.0777917189460475
 MASSIVE-Arabic,spoken-NLU,calendar_query,tiktoken-o200k,125,659,1042,1.5811836115326252,6.029750479846449,1.3074027603513174
-MASSIVE-Arabic,spoken-NLU,calendar_remove,SunAI-64K,67,321,424,1.320872274143302,7.257075471698113,1.0
+MASSIVE-Arabic,spoken-NLU,calendar_remove,SamAI-Tokenizer,67,321,424,1.320872274143302,7.257075471698113,1.0
 MASSIVE-Arabic,spoken-NLU,calendar_remove,Fanar-1,67,321,627,1.9532710280373833,4.907496012759171,1.4787735849056605
 MASSIVE-Arabic,spoken-NLU,calendar_remove,ALLaM-7B,67,321,403,1.2554517133956387,7.635235732009925,0.9504716981132075
 MASSIVE-Arabic,spoken-NLU,calendar_remove,Qwen2.5-7B,67,321,677,2.1090342679127727,4.545051698670606,1.596698113207547
@@ -524,7 +524,7 @@ MASSIVE-Arabic,spoken-NLU,calendar_remove,Mistral-Small-3.1,67,321,588,1.8317757
 MASSIVE-Arabic,spoken-NLU,calendar_remove,Falcon-H1-7B,67,321,552,1.719626168224299,5.574275362318841,1.3018867924528301
 MASSIVE-Arabic,spoken-NLU,calendar_remove,tiktoken-cl100k,67,321,1187,3.6978193146417446,2.5922493681550125,2.7995283018867925
 MASSIVE-Arabic,spoken-NLU,calendar_remove,tiktoken-o200k,67,321,552,1.719626168224299,5.574275362318841,1.3018867924528301
-MASSIVE-Arabic,spoken-NLU,calendar_set,SunAI-64K,205,1362,1739,1.276798825256975,7.5951696377228295,1.0
+MASSIVE-Arabic,spoken-NLU,calendar_set,SamAI-Tokenizer,205,1362,1739,1.276798825256975,7.5951696377228295,1.0
 MASSIVE-Arabic,spoken-NLU,calendar_set,Fanar-1,205,1362,2565,1.8832599118942732,5.149317738791423,1.4749856239217942
 MASSIVE-Arabic,spoken-NLU,calendar_set,ALLaM-7B,205,1362,1653,1.2136563876651982,7.990320629159105,0.9505462909718229
 MASSIVE-Arabic,spoken-NLU,calendar_set,Qwen2.5-7B,205,1362,2697,1.9801762114537445,4.897293288839451,1.5508913168487637
@@ -539,7 +539,7 @@ MASSIVE-Arabic,spoken-NLU,calendar_set,Mistral-Small-3.1,205,1362,2223,1.6321585
 MASSIVE-Arabic,spoken-NLU,calendar_set,Falcon-H1-7B,205,1362,2108,1.5477239353891337,6.265654648956357,1.2121909143185738
 MASSIVE-Arabic,spoken-NLU,calendar_set,tiktoken-cl100k,205,1362,5283,3.8788546255506606,2.5000946431951543,3.037952846463485
 MASSIVE-Arabic,spoken-NLU,calendar_set,tiktoken-o200k,205,1362,2315,1.6997063142437592,5.705399568034557,1.3312248418631398
-MASSIVE-Arabic,spoken-NLU,cooking_recipe,SunAI-64K,72,380,527,1.3868421052631579,6.679316888045541,1.0
+MASSIVE-Arabic,spoken-NLU,cooking_recipe,SamAI-Tokenizer,72,380,527,1.3868421052631579,6.679316888045541,1.0
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,Fanar-1,72,380,731,1.9236842105263159,4.8153214774281805,1.3870967741935485
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,ALLaM-7B,72,380,507,1.3342105263157895,6.942800788954635,0.9620493358633776
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,Qwen2.5-7B,72,380,810,2.1315789473684212,4.345679012345679,1.537001897533207
@@ -554,7 +554,7 @@ MASSIVE-Arabic,spoken-NLU,cooking_recipe,Mistral-Small-3.1,72,380,684,1.8,5.1461
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,Falcon-H1-7B,72,380,663,1.7447368421052631,5.30920060331825,1.2580645161290323
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,tiktoken-cl100k,72,380,1418,3.731578947368421,2.4823695345557124,2.6907020872865277
 MASSIVE-Arabic,spoken-NLU,cooking_recipe,tiktoken-o200k,72,380,698,1.8368421052631578,5.042979942693409,1.3244781783681214
-MASSIVE-Arabic,spoken-NLU,datetime_convert,SunAI-64K,15,111,120,1.0810810810810811,8.433333333333334,1.0
+MASSIVE-Arabic,spoken-NLU,datetime_convert,SamAI-Tokenizer,15,111,120,1.0810810810810811,8.433333333333334,1.0
 MASSIVE-Arabic,spoken-NLU,datetime_convert,Fanar-1,15,111,182,1.6396396396396395,5.56043956043956,1.5166666666666666
 MASSIVE-Arabic,spoken-NLU,datetime_convert,ALLaM-7B,15,111,117,1.054054054054054,8.649572649572649,0.975
 MASSIVE-Arabic,spoken-NLU,datetime_convert,Qwen2.5-7B,15,111,206,1.8558558558558558,4.912621359223301,1.7166666666666666
@@ -569,7 +569,7 @@ MASSIVE-Arabic,spoken-NLU,datetime_convert,Mistral-Small-3.1,15,111,159,1.432432
 MASSIVE-Arabic,spoken-NLU,datetime_convert,Falcon-H1-7B,15,111,148,1.3333333333333333,6.837837837837838,1.2333333333333334
 MASSIVE-Arabic,spoken-NLU,datetime_convert,tiktoken-cl100k,15,111,397,3.5765765765765765,2.5491183879093198,3.308333333333333
 MASSIVE-Arabic,spoken-NLU,datetime_convert,tiktoken-o200k,15,111,163,1.4684684684684686,6.208588957055214,1.3583333333333334
-MASSIVE-Arabic,spoken-NLU,datetime_query,SunAI-64K,84,469,517,1.1023454157782515,7.901353965183753,1.0
+MASSIVE-Arabic,spoken-NLU,datetime_query,SamAI-Tokenizer,84,469,517,1.1023454157782515,7.901353965183753,1.0
 MASSIVE-Arabic,spoken-NLU,datetime_query,Fanar-1,84,469,646,1.3773987206823028,6.323529411764706,1.2495164410058026
 MASSIVE-Arabic,spoken-NLU,datetime_query,ALLaM-7B,84,469,499,1.0639658848614073,8.186372745490981,0.965183752417795
 MASSIVE-Arabic,spoken-NLU,datetime_query,Qwen2.5-7B,84,469,691,1.4733475479744136,5.911722141823445,1.3365570599613152
@@ -584,7 +584,7 @@ MASSIVE-Arabic,spoken-NLU,datetime_query,Mistral-Small-3.1,84,469,602,1.28358208
 MASSIVE-Arabic,spoken-NLU,datetime_query,Falcon-H1-7B,84,469,597,1.2729211087420043,6.842546063651591,1.1547388781431334
 MASSIVE-Arabic,spoken-NLU,datetime_query,tiktoken-cl100k,84,469,1553,3.3113006396588487,2.6303927881519638,3.003868471953578
 MASSIVE-Arabic,spoken-NLU,datetime_query,tiktoken-o200k,84,469,611,1.302771855010661,6.685761047463175,1.1818181818181819
-MASSIVE-Arabic,spoken-NLU,email_addcontact,SunAI-64K,12,95,114,1.2,7.245614035087719,1.0
+MASSIVE-Arabic,spoken-NLU,email_addcontact,SamAI-Tokenizer,12,95,114,1.2,7.245614035087719,1.0
 MASSIVE-Arabic,spoken-NLU,email_addcontact,Fanar-1,12,95,165,1.736842105263158,5.006060606060606,1.4473684210526316
 MASSIVE-Arabic,spoken-NLU,email_addcontact,ALLaM-7B,12,95,114,1.2,7.245614035087719,1.0
 MASSIVE-Arabic,spoken-NLU,email_addcontact,Qwen2.5-7B,12,95,179,1.8842105263157896,4.614525139664805,1.5701754385964912
@@ -599,7 +599,7 @@ MASSIVE-Arabic,spoken-NLU,email_addcontact,Mistral-Small-3.1,12,95,157,1.6526315
 MASSIVE-Arabic,spoken-NLU,email_addcontact,Falcon-H1-7B,12,95,143,1.5052631578947369,5.776223776223776,1.2543859649122806
 MASSIVE-Arabic,spoken-NLU,email_addcontact,tiktoken-cl100k,12,95,297,3.126315789473684,2.781144781144781,2.6052631578947367
 MASSIVE-Arabic,spoken-NLU,email_addcontact,tiktoken-o200k,12,95,155,1.631578947368421,5.329032258064516,1.3596491228070176
-MASSIVE-Arabic,spoken-NLU,email_query,SunAI-64K,118,752,968,1.2872340425531914,6.946280991735537,1.0
+MASSIVE-Arabic,spoken-NLU,email_query,SamAI-Tokenizer,118,752,968,1.2872340425531914,6.946280991735537,1.0
 MASSIVE-Arabic,spoken-NLU,email_query,Fanar-1,118,752,1242,1.6515957446808511,5.413848631239936,1.2830578512396693
 MASSIVE-Arabic,spoken-NLU,email_query,ALLaM-7B,118,752,927,1.2327127659574468,7.253505933117584,0.9576446280991735
 MASSIVE-Arabic,spoken-NLU,email_query,Qwen2.5-7B,118,752,1386,1.8430851063829787,4.851370851370851,1.4318181818181819
@@ -614,7 +614,7 @@ MASSIVE-Arabic,spoken-NLU,email_query,Mistral-Small-3.1,118,752,1227,1.631648936
 MASSIVE-Arabic,spoken-NLU,email_query,Falcon-H1-7B,118,752,1171,1.5571808510638299,5.742100768573868,1.209710743801653
 MASSIVE-Arabic,spoken-NLU,email_query,tiktoken-cl100k,118,752,2526,3.359042553191489,2.6619160728424385,2.6095041322314048
 MASSIVE-Arabic,spoken-NLU,email_query,tiktoken-o200k,118,752,1169,1.5545212765957446,5.751924721984603,1.2076446280991735
-MASSIVE-Arabic,spoken-NLU,email_querycontact,SunAI-64K,26,152,175,1.1513157894736843,7.457142857142857,1.0
+MASSIVE-Arabic,spoken-NLU,email_querycontact,SamAI-Tokenizer,26,152,175,1.1513157894736843,7.457142857142857,1.0
 MASSIVE-Arabic,spoken-NLU,email_querycontact,Fanar-1,26,152,245,1.611842105263158,5.326530612244898,1.4
 MASSIVE-Arabic,spoken-NLU,email_querycontact,ALLaM-7B,26,152,165,1.0855263157894737,7.909090909090909,0.9428571428571428
 MASSIVE-Arabic,spoken-NLU,email_querycontact,Qwen2.5-7B,26,152,277,1.8223684210526316,4.7111913357400725,1.582857142857143
@@ -629,7 +629,7 @@ MASSIVE-Arabic,spoken-NLU,email_querycontact,Mistral-Small-3.1,26,152,241,1.5855
 MASSIVE-Arabic,spoken-NLU,email_querycontact,Falcon-H1-7B,26,152,221,1.4539473684210527,5.904977375565611,1.262857142857143
 MASSIVE-Arabic,spoken-NLU,email_querycontact,tiktoken-cl100k,26,152,512,3.3684210526315788,2.548828125,2.9257142857142857
 MASSIVE-Arabic,spoken-NLU,email_querycontact,tiktoken-o200k,26,152,238,1.5657894736842106,5.483193277310924,1.36
-MASSIVE-Arabic,spoken-NLU,email_sendemail,SunAI-64K,113,749,982,1.3110814419225634,6.773930753564155,1.0
+MASSIVE-Arabic,spoken-NLU,email_sendemail,SamAI-Tokenizer,113,749,982,1.3110814419225634,6.773930753564155,1.0
 MASSIVE-Arabic,spoken-NLU,email_sendemail,Fanar-1,113,749,1390,1.855807743658211,4.785611510791367,1.415478615071283
 MASSIVE-Arabic,spoken-NLU,email_sendemail,ALLaM-7B,113,749,930,1.2416555407209613,7.15268817204301,0.9470468431771895
 MASSIVE-Arabic,spoken-NLU,email_sendemail,Qwen2.5-7B,113,749,1438,1.9198931909212282,4.62586926286509,1.464358452138493
@@ -644,7 +644,7 @@ MASSIVE-Arabic,spoken-NLU,email_sendemail,Mistral-Small-3.1,113,749,1274,1.70093
 MASSIVE-Arabic,spoken-NLU,email_sendemail,Falcon-H1-7B,113,749,1279,1.7076101468624834,5.200938232994527,1.3024439918533606
 MASSIVE-Arabic,spoken-NLU,email_sendemail,tiktoken-cl100k,113,749,2504,3.343124165554072,2.6565495207667733,2.54989816700611
 MASSIVE-Arabic,spoken-NLU,email_sendemail,tiktoken-o200k,113,749,1274,1.7009345794392523,5.221350078492936,1.2973523421588595
-MASSIVE-Arabic,spoken-NLU,general_greet,SunAI-64K,1,2,4,2.0,4.25,1.0
+MASSIVE-Arabic,spoken-NLU,general_greet,SamAI-Tokenizer,1,2,4,2.0,4.25,1.0
 MASSIVE-Arabic,spoken-NLU,general_greet,Fanar-1,1,2,4,2.0,4.25,1.0
 MASSIVE-Arabic,spoken-NLU,general_greet,ALLaM-7B,1,2,2,1.0,8.5,0.5
 MASSIVE-Arabic,spoken-NLU,general_greet,Qwen2.5-7B,1,2,4,2.0,4.25,1.0
@@ -659,7 +659,7 @@ MASSIVE-Arabic,spoken-NLU,general_greet,Mistral-Small-3.1,1,2,4,2.0,4.25,1.0
 MASSIVE-Arabic,spoken-NLU,general_greet,Falcon-H1-7B,1,2,4,2.0,4.25,1.0
 MASSIVE-Arabic,spoken-NLU,general_greet,tiktoken-cl100k,1,2,7,3.5,2.4285714285714284,1.75
 MASSIVE-Arabic,spoken-NLU,general_greet,tiktoken-o200k,1,2,4,2.0,4.25,1.0
-MASSIVE-Arabic,spoken-NLU,general_joke,SunAI-64K,19,66,111,1.6818181818181819,5.126126126126126,1.0
+MASSIVE-Arabic,spoken-NLU,general_joke,SamAI-Tokenizer,19,66,111,1.6818181818181819,5.126126126126126,1.0
 MASSIVE-Arabic,spoken-NLU,general_joke,Fanar-1,19,66,142,2.1515151515151514,4.007042253521127,1.2792792792792793
 MASSIVE-Arabic,spoken-NLU,general_joke,ALLaM-7B,19,66,92,1.393939393939394,6.184782608695652,0.8288288288288288
 MASSIVE-Arabic,spoken-NLU,general_joke,Qwen2.5-7B,19,66,151,2.287878787878788,3.7682119205298013,1.3603603603603605
@@ -674,7 +674,7 @@ MASSIVE-Arabic,spoken-NLU,general_joke,Mistral-Small-3.1,19,66,139,2.10606060606
 MASSIVE-Arabic,spoken-NLU,general_joke,Falcon-H1-7B,19,66,140,2.121212121212121,4.064285714285714,1.2612612612612613
 MASSIVE-Arabic,spoken-NLU,general_joke,tiktoken-cl100k,19,66,244,3.696969696969697,2.331967213114754,2.1981981981981984
 MASSIVE-Arabic,spoken-NLU,general_joke,tiktoken-o200k,19,66,143,2.1666666666666665,3.979020979020979,1.2882882882882882
-MASSIVE-Arabic,spoken-NLU,general_quirky,SunAI-64K,169,889,1149,1.2924634420697412,7.2645778938207135,1.0
+MASSIVE-Arabic,spoken-NLU,general_quirky,SamAI-Tokenizer,169,889,1149,1.2924634420697412,7.2645778938207135,1.0
 MASSIVE-Arabic,spoken-NLU,general_quirky,Fanar-1,169,889,1560,1.7547806524184477,5.350641025641026,1.3577023498694516
 MASSIVE-Arabic,spoken-NLU,general_quirky,ALLaM-7B,169,889,1061,1.1934758155230596,7.867106503298775,0.9234116623150566
 MASSIVE-Arabic,spoken-NLU,general_quirky,Qwen2.5-7B,169,889,1728,1.9437570303712035,4.830439814814815,1.5039164490861618
@@ -689,7 +689,7 @@ MASSIVE-Arabic,spoken-NLU,general_quirky,Mistral-Small-3.1,169,889,1439,1.618672
 MASSIVE-Arabic,spoken-NLU,general_quirky,Falcon-H1-7B,169,889,1403,1.578177727784027,5.949394155381325,1.2210617928633594
 MASSIVE-Arabic,spoken-NLU,general_quirky,tiktoken-cl100k,169,889,3305,3.717660292463442,2.5255673222390316,2.876414273281114
 MASSIVE-Arabic,spoken-NLU,general_quirky,tiktoken-o200k,169,889,1472,1.6557930258717661,5.670516304347826,1.2811140121845084
-MASSIVE-Arabic,spoken-NLU,iot_cleaning,SunAI-64K,21,79,104,1.3164556962025316,8.10576923076923,1.0
+MASSIVE-Arabic,spoken-NLU,iot_cleaning,SamAI-Tokenizer,21,79,104,1.3164556962025316,8.10576923076923,1.0
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,Fanar-1,21,79,166,2.1012658227848102,5.078313253012048,1.5961538461538463
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,ALLaM-7B,21,79,102,1.2911392405063291,8.264705882352942,0.9807692307692307
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,Qwen2.5-7B,21,79,183,2.3164556962025316,4.60655737704918,1.7596153846153846
@@ -704,7 +704,7 @@ MASSIVE-Arabic,spoken-NLU,iot_cleaning,Mistral-Small-3.1,21,79,153,1.93670886075
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,Falcon-H1-7B,21,79,151,1.9113924050632911,5.582781456953643,1.4519230769230769
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,tiktoken-cl100k,21,79,334,4.227848101265823,2.5239520958083834,3.2115384615384617
 MASSIVE-Arabic,spoken-NLU,iot_cleaning,tiktoken-o200k,21,79,156,1.9746835443037976,5.403846153846154,1.5
-MASSIVE-Arabic,spoken-NLU,iot_coffee,SunAI-64K,35,150,196,1.3066666666666666,6.780612244897959,1.0
+MASSIVE-Arabic,spoken-NLU,iot_coffee,SamAI-Tokenizer,35,150,196,1.3066666666666666,6.780612244897959,1.0
 MASSIVE-Arabic,spoken-NLU,iot_coffee,Fanar-1,35,150,292,1.9466666666666668,4.551369863013699,1.489795918367347
 MASSIVE-Arabic,spoken-NLU,iot_coffee,ALLaM-7B,35,150,178,1.1866666666666668,7.466292134831461,0.9081632653061225
 MASSIVE-Arabic,spoken-NLU,iot_coffee,Qwen2.5-7B,35,150,326,2.1733333333333333,4.076687116564417,1.663265306122449
@@ -719,7 +719,7 @@ MASSIVE-Arabic,spoken-NLU,iot_coffee,Mistral-Small-3.1,35,150,267,1.78,4.9775280
 MASSIVE-Arabic,spoken-NLU,iot_coffee,Falcon-H1-7B,35,150,266,1.7733333333333334,4.996240601503759,1.3571428571428572
 MASSIVE-Arabic,spoken-NLU,iot_coffee,tiktoken-cl100k,35,150,535,3.566666666666667,2.4841121495327103,2.729591836734694
 MASSIVE-Arabic,spoken-NLU,iot_coffee,tiktoken-o200k,35,150,274,1.8266666666666667,4.850364963503649,1.3979591836734695
-MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,SunAI-64K,36,201,265,1.318407960199005,6.667924528301887,1.0
+MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,SamAI-Tokenizer,36,201,265,1.318407960199005,6.667924528301887,1.0
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,Fanar-1,36,201,357,1.7761194029850746,4.949579831932773,1.3471698113207546
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,ALLaM-7B,36,201,256,1.2736318407960199,6.90234375,0.9660377358490566
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,Qwen2.5-7B,36,201,389,1.935323383084577,4.5424164524421595,1.4679245283018867
@@ -734,7 +734,7 @@ MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,Mistral-Small-3.1,36,201,333,1.656
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,Falcon-H1-7B,36,201,339,1.6865671641791045,5.212389380530974,1.2792452830188679
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,tiktoken-cl100k,36,201,648,3.2238805970149254,2.7268518518518516,2.4452830188679244
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightchange,tiktoken-o200k,36,201,341,1.6965174129353233,5.181818181818182,1.2867924528301886
-MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,SunAI-64K,21,89,122,1.3707865168539326,6.737704918032787,1.0
+MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,SamAI-Tokenizer,21,89,122,1.3707865168539326,6.737704918032787,1.0
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,Fanar-1,21,89,184,2.067415730337079,4.467391304347826,1.5081967213114753
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,ALLaM-7B,21,89,130,1.4606741573033708,6.323076923076923,1.0655737704918034
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,Qwen2.5-7B,21,89,192,2.157303370786517,4.28125,1.5737704918032787
@@ -749,7 +749,7 @@ MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,Mistral-Small-3.1,21,89,171,1.9213483
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,Falcon-H1-7B,21,89,165,1.853932584269663,4.9818181818181815,1.3524590163934427
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,tiktoken-cl100k,21,89,313,3.5168539325842696,2.626198083067093,2.5655737704918034
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightdim,tiktoken-o200k,21,89,179,2.0112359550561796,4.592178770949721,1.4672131147540983
-MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,SunAI-64K,40,158,205,1.2974683544303798,6.902439024390244,1.0
+MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,SamAI-Tokenizer,40,158,205,1.2974683544303798,6.902439024390244,1.0
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,Fanar-1,40,158,287,1.8164556962025316,4.930313588850174,1.4
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,ALLaM-7B,40,158,216,1.3670886075949367,6.550925925925926,1.053658536585366
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,Qwen2.5-7B,40,158,306,1.9367088607594938,4.624183006535947,1.4926829268292683
@@ -764,7 +764,7 @@ MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,Mistral-Small-3.1,40,158,269,1.702531
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,Falcon-H1-7B,40,158,273,1.7278481012658229,5.183150183150183,1.3317073170731708
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,tiktoken-cl100k,40,158,531,3.3607594936708862,2.664783427495292,2.5902439024390245
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightoff,tiktoken-o200k,40,158,277,1.7531645569620253,5.108303249097473,1.351219512195122
-MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,SunAI-64K,3,11,17,1.5454545454545454,6.470588235294118,1.0
+MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,SamAI-Tokenizer,3,11,17,1.5454545454545454,6.470588235294118,1.0
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,Fanar-1,3,11,24,2.1818181818181817,4.583333333333333,1.411764705882353
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,ALLaM-7B,3,11,16,1.4545454545454546,6.875,0.9411764705882353
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,Qwen2.5-7B,3,11,23,2.090909090909091,4.782608695652174,1.3529411764705883
@@ -779,7 +779,7 @@ MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,Mistral-Small-3.1,3,11,23,2.0909090909
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,Falcon-H1-7B,3,11,24,2.1818181818181817,4.583333333333333,1.411764705882353
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,tiktoken-cl100k,3,11,44,4.0,2.5,2.588235294117647
 MASSIVE-Arabic,spoken-NLU,iot_hue_lighton,tiktoken-o200k,3,11,23,2.090909090909091,4.782608695652174,1.3529411764705883
-MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,SunAI-64K,27,109,141,1.2935779816513762,6.929078014184397,1.0
+MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,SamAI-Tokenizer,27,109,141,1.2935779816513762,6.929078014184397,1.0
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,Fanar-1,27,109,201,1.8440366972477065,4.8606965174129355,1.425531914893617
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,ALLaM-7B,27,109,127,1.165137614678899,7.692913385826771,0.900709219858156
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,Qwen2.5-7B,27,109,217,1.9908256880733946,4.502304147465438,1.5390070921985815
@@ -794,7 +794,7 @@ MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,Mistral-Small-3.1,27,109,187,1.7155963
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,Falcon-H1-7B,27,109,182,1.6697247706422018,5.368131868131868,1.2907801418439717
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,tiktoken-cl100k,27,109,372,3.4128440366972477,2.6263440860215055,2.6382978723404253
 MASSIVE-Arabic,spoken-NLU,iot_hue_lightup,tiktoken-o200k,27,109,194,1.7798165137614679,5.036082474226804,1.375886524822695
-MASSIVE-Arabic,spoken-NLU,iot_wemo_off,SunAI-64K,18,62,88,1.4193548387096775,6.113636363636363,1.0
+MASSIVE-Arabic,spoken-NLU,iot_wemo_off,SamAI-Tokenizer,18,62,88,1.4193548387096775,6.113636363636363,1.0
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,Fanar-1,18,62,121,1.9516129032258065,4.446280991735537,1.375
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,ALLaM-7B,18,62,96,1.5483870967741935,5.604166666666667,1.0909090909090908
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,Qwen2.5-7B,18,62,135,2.1774193548387095,3.9851851851851854,1.5340909090909092
@@ -809,7 +809,7 @@ MASSIVE-Arabic,spoken-NLU,iot_wemo_off,Mistral-Small-3.1,18,62,113,1.82258064516
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,Falcon-H1-7B,18,62,110,1.7741935483870968,4.890909090909091,1.25
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,tiktoken-cl100k,18,62,216,3.4838709677419355,2.490740740740741,2.4545454545454546
 MASSIVE-Arabic,spoken-NLU,iot_wemo_off,tiktoken-o200k,18,62,115,1.8548387096774193,4.678260869565217,1.3068181818181819
-MASSIVE-Arabic,spoken-NLU,iot_wemo_on,SunAI-64K,8,31,39,1.2580645161290323,6.641025641025641,1.0
+MASSIVE-Arabic,spoken-NLU,iot_wemo_on,SamAI-Tokenizer,8,31,39,1.2580645161290323,6.641025641025641,1.0
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,Fanar-1,8,31,55,1.7741935483870968,4.709090909090909,1.4102564102564104
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,ALLaM-7B,8,31,38,1.2258064516129032,6.815789473684211,0.9743589743589743
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,Qwen2.5-7B,8,31,53,1.7096774193548387,4.886792452830188,1.358974358974359
@@ -824,7 +824,7 @@ MASSIVE-Arabic,spoken-NLU,iot_wemo_on,Mistral-Small-3.1,8,31,51,1.64516129032258
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,Falcon-H1-7B,8,31,51,1.6451612903225807,5.078431372549019,1.3076923076923077
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,tiktoken-cl100k,8,31,102,3.2903225806451615,2.5392156862745097,2.6153846153846154
 MASSIVE-Arabic,spoken-NLU,iot_wemo_on,tiktoken-o200k,8,31,51,1.6451612903225807,5.078431372549019,1.3076923076923077
-MASSIVE-Arabic,spoken-NLU,lists_createoradd,SunAI-64K,38,193,230,1.1917098445595855,8.404347826086957,1.0
+MASSIVE-Arabic,spoken-NLU,lists_createoradd,SamAI-Tokenizer,38,193,230,1.1917098445595855,8.404347826086957,1.0
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,Fanar-1,38,193,360,1.8652849740932642,5.3694444444444445,1.565217391304348
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,ALLaM-7B,38,193,251,1.3005181347150259,7.701195219123506,1.0913043478260869
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,Qwen2.5-7B,38,193,429,2.22279792746114,4.505827505827506,1.8652173913043477
@@ -839,7 +839,7 @@ MASSIVE-Arabic,spoken-NLU,lists_createoradd,Mistral-Small-3.1,38,193,327,1.69430
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,Falcon-H1-7B,38,193,321,1.6632124352331605,6.021806853582555,1.3956521739130434
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,tiktoken-cl100k,38,193,786,4.072538860103627,2.459287531806616,3.417391304347826
 MASSIVE-Arabic,spoken-NLU,lists_createoradd,tiktoken-o200k,38,193,335,1.7357512953367875,5.7701492537313435,1.4565217391304348
-MASSIVE-Arabic,spoken-NLU,lists_query,SunAI-64K,51,240,281,1.1708333333333334,8.491103202846976,1.0
+MASSIVE-Arabic,spoken-NLU,lists_query,SamAI-Tokenizer,51,240,281,1.1708333333333334,8.491103202846976,1.0
 MASSIVE-Arabic,spoken-NLU,lists_query,Fanar-1,51,240,433,1.8041666666666667,5.510392609699769,1.5409252669039146
 MASSIVE-Arabic,spoken-NLU,lists_query,ALLaM-7B,51,240,268,1.1166666666666667,8.902985074626866,0.9537366548042705
 MASSIVE-Arabic,spoken-NLU,lists_query,Qwen2.5-7B,51,240,494,2.058333333333333,4.82995951417004,1.7580071174377223
@@ -854,7 +854,7 @@ MASSIVE-Arabic,spoken-NLU,lists_query,Mistral-Small-3.1,51,240,385,1.60416666666
 MASSIVE-Arabic,spoken-NLU,lists_query,Falcon-H1-7B,51,240,379,1.5791666666666666,6.295514511873351,1.3487544483985765
 MASSIVE-Arabic,spoken-NLU,lists_query,tiktoken-cl100k,51,240,915,3.8125,2.607650273224044,3.2562277580071175
 MASSIVE-Arabic,spoken-NLU,lists_query,tiktoken-o200k,51,240,391,1.6291666666666667,6.10230179028133,1.3914590747330962
-MASSIVE-Arabic,spoken-NLU,lists_remove,SunAI-64K,49,223,298,1.336322869955157,7.3624161073825505,1.0
+MASSIVE-Arabic,spoken-NLU,lists_remove,SamAI-Tokenizer,49,223,298,1.336322869955157,7.3624161073825505,1.0
 MASSIVE-Arabic,spoken-NLU,lists_remove,Fanar-1,49,223,436,1.9551569506726458,5.032110091743119,1.4630872483221478
 MASSIVE-Arabic,spoken-NLU,lists_remove,ALLaM-7B,49,223,288,1.2914798206278026,7.618055555555555,0.9664429530201343
 MASSIVE-Arabic,spoken-NLU,lists_remove,Qwen2.5-7B,49,223,505,2.2645739910313902,4.344554455445545,1.6946308724832215
@@ -869,7 +869,7 @@ MASSIVE-Arabic,spoken-NLU,lists_remove,Mistral-Small-3.1,49,223,409,1.8340807174
 MASSIVE-Arabic,spoken-NLU,lists_remove,Falcon-H1-7B,49,223,397,1.7802690582959642,5.526448362720403,1.332214765100671
 MASSIVE-Arabic,spoken-NLU,lists_remove,tiktoken-cl100k,49,223,851,3.8161434977578477,2.5781433607520565,2.8557046979865772
 MASSIVE-Arabic,spoken-NLU,lists_remove,tiktoken-o200k,49,223,393,1.7623318385650224,5.582697201017812,1.3187919463087248
-MASSIVE-Arabic,spoken-NLU,music_dislikeness,SunAI-64K,4,18,21,1.1666666666666667,6.142857142857143,1.0
+MASSIVE-Arabic,spoken-NLU,music_dislikeness,SamAI-Tokenizer,4,18,21,1.1666666666666667,6.142857142857143,1.0
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,Fanar-1,4,18,30,1.6666666666666667,4.3,1.4285714285714286
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,ALLaM-7B,4,18,18,1.0,7.166666666666667,0.8571428571428571
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,Qwen2.5-7B,4,18,32,1.7777777777777777,4.03125,1.5238095238095237
@@ -884,7 +884,7 @@ MASSIVE-Arabic,spoken-NLU,music_dislikeness,Mistral-Small-3.1,4,18,24,1.33333333
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,Falcon-H1-7B,4,18,25,1.3888888888888888,5.16,1.1904761904761905
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,tiktoken-cl100k,4,18,47,2.611111111111111,2.74468085106383,2.238095238095238
 MASSIVE-Arabic,spoken-NLU,music_dislikeness,tiktoken-o200k,4,18,29,1.6111111111111112,4.448275862068965,1.380952380952381
-MASSIVE-Arabic,spoken-NLU,music_likeness,SunAI-64K,36,165,211,1.2787878787878788,7.208530805687204,1.0
+MASSIVE-Arabic,spoken-NLU,music_likeness,SamAI-Tokenizer,36,165,211,1.2787878787878788,7.208530805687204,1.0
 MASSIVE-Arabic,spoken-NLU,music_likeness,Fanar-1,36,165,314,1.903030303030303,4.843949044585988,1.4881516587677726
 MASSIVE-Arabic,spoken-NLU,music_likeness,ALLaM-7B,36,165,181,1.096969696969697,8.403314917127071,0.8578199052132701
 MASSIVE-Arabic,spoken-NLU,music_likeness,Qwen2.5-7B,36,165,360,2.1818181818181817,4.225,1.7061611374407584
@@ -899,7 +899,7 @@ MASSIVE-Arabic,spoken-NLU,music_likeness,Mistral-Small-3.1,36,165,267,1.61818181
 MASSIVE-Arabic,spoken-NLU,music_likeness,Falcon-H1-7B,36,165,266,1.612121212121212,5.7180451127819545,1.2606635071090047
 MASSIVE-Arabic,spoken-NLU,music_likeness,tiktoken-cl100k,36,165,608,3.684848484848485,2.5016447368421053,2.881516587677725
 MASSIVE-Arabic,spoken-NLU,music_likeness,tiktoken-o200k,36,165,289,1.7515151515151515,5.262975778546712,1.3696682464454977
-MASSIVE-Arabic,spoken-NLU,music_query,SunAI-64K,33,129,160,1.2403100775193798,7.775,1.0
+MASSIVE-Arabic,spoken-NLU,music_query,SamAI-Tokenizer,33,129,160,1.2403100775193798,7.775,1.0
 MASSIVE-Arabic,spoken-NLU,music_query,Fanar-1,33,129,234,1.813953488372093,5.316239316239316,1.4625
 MASSIVE-Arabic,spoken-NLU,music_query,ALLaM-7B,33,129,143,1.1085271317829457,8.6993006993007,0.89375
 MASSIVE-Arabic,spoken-NLU,music_query,Qwen2.5-7B,33,129,263,2.0387596899224807,4.730038022813688,1.64375
@@ -914,7 +914,7 @@ MASSIVE-Arabic,spoken-NLU,music_query,Mistral-Small-3.1,33,129,195,1.51162790697
 MASSIVE-Arabic,spoken-NLU,music_query,Falcon-H1-7B,33,129,197,1.5271317829457365,6.314720812182741,1.23125
 MASSIVE-Arabic,spoken-NLU,music_query,tiktoken-cl100k,33,129,480,3.7209302325581395,2.591666666666667,3.0
 MASSIVE-Arabic,spoken-NLU,music_query,tiktoken-o200k,33,129,221,1.7131782945736433,5.6289592760181,1.38125
-MASSIVE-Arabic,spoken-NLU,music_settings,SunAI-64K,6,19,24,1.263157894736842,7.458333333333333,1.0
+MASSIVE-Arabic,spoken-NLU,music_settings,SamAI-Tokenizer,6,19,24,1.263157894736842,7.458333333333333,1.0
 MASSIVE-Arabic,spoken-NLU,music_settings,Fanar-1,6,19,37,1.9473684210526316,4.837837837837838,1.5416666666666667
 MASSIVE-Arabic,spoken-NLU,music_settings,ALLaM-7B,6,19,22,1.1578947368421053,8.136363636363637,0.9166666666666666
 MASSIVE-Arabic,spoken-NLU,music_settings,Qwen2.5-7B,6,19,42,2.210526315789474,4.261904761904762,1.75
@@ -929,7 +929,7 @@ MASSIVE-Arabic,spoken-NLU,music_settings,Mistral-Small-3.1,6,19,33,1.73684210526
 MASSIVE-Arabic,spoken-NLU,music_settings,Falcon-H1-7B,6,19,32,1.6842105263157894,5.59375,1.3333333333333333
 MASSIVE-Arabic,spoken-NLU,music_settings,tiktoken-cl100k,6,19,68,3.5789473684210527,2.6323529411764706,2.8333333333333335
 MASSIVE-Arabic,spoken-NLU,music_settings,tiktoken-o200k,6,19,34,1.7894736842105263,5.264705882352941,1.4166666666666667
-MASSIVE-Arabic,spoken-NLU,news_query,SunAI-64K,124,715,800,1.118881118881119,8.69375,1.0
+MASSIVE-Arabic,spoken-NLU,news_query,SamAI-Tokenizer,124,715,800,1.118881118881119,8.69375,1.0
 MASSIVE-Arabic,spoken-NLU,news_query,Fanar-1,124,715,1141,1.5958041958041957,6.095530236634531,1.42625
 MASSIVE-Arabic,spoken-NLU,news_query,ALLaM-7B,124,715,758,1.0601398601398602,9.175461741424803,0.9475
 MASSIVE-Arabic,spoken-NLU,news_query,Qwen2.5-7B,124,715,1233,1.7244755244755245,5.640713706407137,1.54125
@@ -944,7 +944,7 @@ MASSIVE-Arabic,spoken-NLU,news_query,Mistral-Small-3.1,124,715,993,1.38881118881
 MASSIVE-Arabic,spoken-NLU,news_query,Falcon-H1-7B,124,715,923,1.290909090909091,7.535211267605634,1.15375
 MASSIVE-Arabic,spoken-NLU,news_query,tiktoken-cl100k,124,715,2613,3.6545454545454548,2.6616915422885574,3.26625
 MASSIVE-Arabic,spoken-NLU,news_query,tiktoken-o200k,124,715,998,1.3958041958041958,6.968937875751503,1.2475
-MASSIVE-Arabic,spoken-NLU,play_audiobook,SunAI-64K,41,219,280,1.278538812785388,7.203571428571428,1.0
+MASSIVE-Arabic,spoken-NLU,play_audiobook,SamAI-Tokenizer,41,219,280,1.278538812785388,7.203571428571428,1.0
 MASSIVE-Arabic,spoken-NLU,play_audiobook,Fanar-1,41,219,371,1.6940639269406392,5.436657681940701,1.325
 MASSIVE-Arabic,spoken-NLU,play_audiobook,ALLaM-7B,41,219,267,1.2191780821917808,7.5543071161048685,0.9535714285714286
 MASSIVE-Arabic,spoken-NLU,play_audiobook,Qwen2.5-7B,41,219,406,1.8538812785388128,4.967980295566503,1.45
@@ -959,7 +959,7 @@ MASSIVE-Arabic,spoken-NLU,play_audiobook,Mistral-Small-3.1,41,219,339,1.54794520
 MASSIVE-Arabic,spoken-NLU,play_audiobook,Falcon-H1-7B,41,219,332,1.5159817351598173,6.075301204819277,1.1857142857142857
 MASSIVE-Arabic,spoken-NLU,play_audiobook,tiktoken-cl100k,41,219,732,3.3424657534246576,2.7554644808743167,2.6142857142857143
 MASSIVE-Arabic,spoken-NLU,play_audiobook,tiktoken-o200k,41,219,351,1.6027397260273972,5.746438746438747,1.2535714285714286
-MASSIVE-Arabic,spoken-NLU,play_game,SunAI-64K,34,106,148,1.3962264150943395,5.993243243243243,1.0
+MASSIVE-Arabic,spoken-NLU,play_game,SamAI-Tokenizer,34,106,148,1.3962264150943395,5.993243243243243,1.0
 MASSIVE-Arabic,spoken-NLU,play_game,Fanar-1,34,106,194,1.830188679245283,4.572164948453608,1.3108108108108107
 MASSIVE-Arabic,spoken-NLU,play_game,ALLaM-7B,34,106,144,1.3584905660377358,6.159722222222222,0.972972972972973
 MASSIVE-Arabic,spoken-NLU,play_game,Qwen2.5-7B,34,106,207,1.9528301886792452,4.285024154589372,1.3986486486486487
@@ -974,7 +974,7 @@ MASSIVE-Arabic,spoken-NLU,play_game,Mistral-Small-3.1,34,106,189,1.7830188679245
 MASSIVE-Arabic,spoken-NLU,play_game,Falcon-H1-7B,34,106,190,1.7924528301886793,4.668421052631579,1.2837837837837838
 MASSIVE-Arabic,spoken-NLU,play_game,tiktoken-cl100k,34,106,330,3.1132075471698113,2.687878787878788,2.22972972972973
 MASSIVE-Arabic,spoken-NLU,play_game,tiktoken-o200k,34,106,185,1.7452830188679245,4.794594594594595,1.25
-MASSIVE-Arabic,spoken-NLU,play_music,SunAI-64K,174,826,1051,1.2723970944309928,6.95337773549001,1.0
+MASSIVE-Arabic,spoken-NLU,play_music,SamAI-Tokenizer,174,826,1051,1.2723970944309928,6.95337773549001,1.0
 MASSIVE-Arabic,spoken-NLU,play_music,Fanar-1,174,826,1512,1.8305084745762712,4.833333333333333,1.4386298763082779
 MASSIVE-Arabic,spoken-NLU,play_music,ALLaM-7B,174,826,982,1.188861985472155,7.441955193482689,0.934348239771646
 MASSIVE-Arabic,spoken-NLU,play_music,Qwen2.5-7B,174,826,1644,1.9903147699757868,4.445255474452555,1.5642245480494767
@@ -989,7 +989,7 @@ MASSIVE-Arabic,spoken-NLU,play_music,Mistral-Small-3.1,174,826,1391,1.6840193704
 MASSIVE-Arabic,spoken-NLU,play_music,Falcon-H1-7B,174,826,1345,1.6283292978208233,5.433457249070632,1.279733587059943
 MASSIVE-Arabic,spoken-NLU,play_music,tiktoken-cl100k,174,826,2778,3.36319612590799,2.6306695464362853,2.643196955280685
 MASSIVE-Arabic,spoken-NLU,play_music,tiktoken-o200k,174,826,1438,1.7409200968523002,5.082058414464534,1.368220742150333
-MASSIVE-Arabic,spoken-NLU,play_podcasts,SunAI-64K,63,295,379,1.2847457627118644,7.382585751978892,1.0
+MASSIVE-Arabic,spoken-NLU,play_podcasts,SamAI-Tokenizer,63,295,379,1.2847457627118644,7.382585751978892,1.0
 MASSIVE-Arabic,spoken-NLU,play_podcasts,Fanar-1,63,295,530,1.7966101694915255,5.279245283018868,1.3984168865435356
 MASSIVE-Arabic,spoken-NLU,play_podcasts,ALLaM-7B,63,295,365,1.2372881355932204,7.6657534246575345,0.9630606860158312
 MASSIVE-Arabic,spoken-NLU,play_podcasts,Qwen2.5-7B,63,295,574,1.9457627118644067,4.874564459930314,1.5145118733509235
@@ -1004,7 +1004,7 @@ MASSIVE-Arabic,spoken-NLU,play_podcasts,Mistral-Small-3.1,63,295,501,1.698305084
 MASSIVE-Arabic,spoken-NLU,play_podcasts,Falcon-H1-7B,63,295,482,1.6338983050847458,5.804979253112033,1.2717678100263852
 MASSIVE-Arabic,spoken-NLU,play_podcasts,tiktoken-cl100k,63,295,989,3.352542372881356,2.8291203235591507,2.609498680738786
 MASSIVE-Arabic,spoken-NLU,play_podcasts,tiktoken-o200k,63,295,496,1.6813559322033897,5.641129032258065,1.3087071240105541
-MASSIVE-Arabic,spoken-NLU,play_radio,SunAI-64K,67,309,401,1.2977346278317152,6.211970074812967,1.0
+MASSIVE-Arabic,spoken-NLU,play_radio,SamAI-Tokenizer,67,309,401,1.2977346278317152,6.211970074812967,1.0
 MASSIVE-Arabic,spoken-NLU,play_radio,Fanar-1,67,309,527,1.7055016181229774,4.726755218216319,1.314214463840399
 MASSIVE-Arabic,spoken-NLU,play_radio,ALLaM-7B,67,309,364,1.1779935275080906,6.843406593406593,0.9077306733167082
 MASSIVE-Arabic,spoken-NLU,play_radio,Qwen2.5-7B,67,309,557,1.802588996763754,4.472172351885098,1.3890274314214464
@@ -1019,7 +1019,7 @@ MASSIVE-Arabic,spoken-NLU,play_radio,Mistral-Small-3.1,67,309,493,1.595469255663
 MASSIVE-Arabic,spoken-NLU,play_radio,Falcon-H1-7B,67,309,472,1.5275080906148868,5.2775423728813555,1.1770573566084788
 MASSIVE-Arabic,spoken-NLU,play_radio,tiktoken-cl100k,67,309,933,3.0194174757281553,2.669882100750268,2.3266832917705735
 MASSIVE-Arabic,spoken-NLU,play_radio,tiktoken-o200k,67,309,512,1.656957928802589,4.865234375,1.2768079800498753
-MASSIVE-Arabic,spoken-NLU,qa_currency,SunAI-64K,38,253,279,1.1027667984189724,9.229390681003585,1.0
+MASSIVE-Arabic,spoken-NLU,qa_currency,SamAI-Tokenizer,38,253,279,1.1027667984189724,9.229390681003585,1.0
 MASSIVE-Arabic,spoken-NLU,qa_currency,Fanar-1,38,253,406,1.6047430830039526,6.342364532019705,1.4551971326164874
 MASSIVE-Arabic,spoken-NLU,qa_currency,ALLaM-7B,38,253,259,1.023715415019763,9.942084942084943,0.9283154121863799
 MASSIVE-Arabic,spoken-NLU,qa_currency,Qwen2.5-7B,38,253,425,1.6798418972332017,6.0588235294117645,1.5232974910394266
@@ -1034,7 +1034,7 @@ MASSIVE-Arabic,spoken-NLU,qa_currency,Mistral-Small-3.1,38,253,366,1.44664031620
 MASSIVE-Arabic,spoken-NLU,qa_currency,Falcon-H1-7B,38,253,306,1.209486166007905,8.415032679738562,1.096774193548387
 MASSIVE-Arabic,spoken-NLU,qa_currency,tiktoken-cl100k,38,253,939,3.7114624505928853,2.742279020234292,3.3655913978494625
 MASSIVE-Arabic,spoken-NLU,qa_currency,tiktoken-o200k,38,253,360,1.4229249011857708,7.152777777777778,1.2903225806451613
-MASSIVE-Arabic,spoken-NLU,qa_definition,SunAI-64K,57,193,239,1.238341968911917,7.523012552301255,1.0
+MASSIVE-Arabic,spoken-NLU,qa_definition,SamAI-Tokenizer,57,193,239,1.238341968911917,7.523012552301255,1.0
 MASSIVE-Arabic,spoken-NLU,qa_definition,Fanar-1,57,193,351,1.8186528497409327,5.122507122507122,1.4686192468619246
 MASSIVE-Arabic,spoken-NLU,qa_definition,ALLaM-7B,57,193,232,1.2020725388601037,7.75,0.9707112970711297
 MASSIVE-Arabic,spoken-NLU,qa_definition,Qwen2.5-7B,57,193,371,1.922279792746114,4.846361185983827,1.5523012552301256
@@ -1049,7 +1049,7 @@ MASSIVE-Arabic,spoken-NLU,qa_definition,Mistral-Small-3.1,57,193,324,1.678756476
 MASSIVE-Arabic,spoken-NLU,qa_definition,Falcon-H1-7B,57,193,331,1.7150259067357514,5.43202416918429,1.3849372384937237
 MASSIVE-Arabic,spoken-NLU,qa_definition,tiktoken-cl100k,57,193,739,3.8290155440414506,2.4330175913396483,3.092050209205021
 MASSIVE-Arabic,spoken-NLU,qa_definition,tiktoken-o200k,57,193,336,1.7409326424870466,5.351190476190476,1.405857740585774
-MASSIVE-Arabic,spoken-NLU,qa_factoid,SunAI-64K,140,766,921,1.2023498694516972,7.609120521172638,1.0
+MASSIVE-Arabic,spoken-NLU,qa_factoid,SamAI-Tokenizer,140,766,921,1.2023498694516972,7.609120521172638,1.0
 MASSIVE-Arabic,spoken-NLU,qa_factoid,Fanar-1,140,766,1273,1.6618798955613576,5.5051060487038495,1.3821932681867535
 MASSIVE-Arabic,spoken-NLU,qa_factoid,ALLaM-7B,140,766,898,1.1723237597911227,7.804008908685969,0.9750271444082519
 MASSIVE-Arabic,spoken-NLU,qa_factoid,Qwen2.5-7B,140,766,1442,1.8825065274151436,4.859916782246879,1.5656894679695983
@@ -1064,7 +1064,7 @@ MASSIVE-Arabic,spoken-NLU,qa_factoid,Mistral-Small-3.1,140,766,1154,1.5065274151
 MASSIVE-Arabic,spoken-NLU,qa_factoid,Falcon-H1-7B,140,766,1127,1.4712793733681462,6.218278615794143,1.223669923995657
 MASSIVE-Arabic,spoken-NLU,qa_factoid,tiktoken-cl100k,140,766,2708,3.535248041775457,2.587887740029542,2.9402823018458197
 MASSIVE-Arabic,spoken-NLU,qa_factoid,tiktoken-o200k,140,766,1218,1.5900783289817233,5.753694581280788,1.3224755700325732
-MASSIVE-Arabic,spoken-NLU,qa_maths,SunAI-64K,25,160,198,1.2375,7.459595959595959,1.0
+MASSIVE-Arabic,spoken-NLU,qa_maths,SamAI-Tokenizer,25,160,198,1.2375,7.459595959595959,1.0
 MASSIVE-Arabic,spoken-NLU,qa_maths,Fanar-1,25,160,288,1.8,5.128472222222222,1.4545454545454546
 MASSIVE-Arabic,spoken-NLU,qa_maths,ALLaM-7B,25,160,193,1.20625,7.652849740932642,0.9747474747474747
 MASSIVE-Arabic,spoken-NLU,qa_maths,Qwen2.5-7B,25,160,332,2.075,4.448795180722891,1.6767676767676767
@@ -1079,7 +1079,7 @@ MASSIVE-Arabic,spoken-NLU,qa_maths,Mistral-Small-3.1,25,160,254,1.5875,5.8149606
 MASSIVE-Arabic,spoken-NLU,qa_maths,Falcon-H1-7B,25,160,255,1.59375,5.792156862745098,1.2878787878787878
 MASSIVE-Arabic,spoken-NLU,qa_maths,tiktoken-cl100k,25,160,631,3.94375,2.340729001584786,3.186868686868687
 MASSIVE-Arabic,spoken-NLU,qa_maths,tiktoken-o200k,25,160,265,1.65625,5.573584905660377,1.3383838383838385
-MASSIVE-Arabic,spoken-NLU,qa_stock,SunAI-64K,26,142,165,1.1619718309859155,8.181818181818182,1.0
+MASSIVE-Arabic,spoken-NLU,qa_stock,SamAI-Tokenizer,26,142,165,1.1619718309859155,8.181818181818182,1.0
 MASSIVE-Arabic,spoken-NLU,qa_stock,Fanar-1,26,142,264,1.8591549295774648,5.113636363636363,1.6
 MASSIVE-Arabic,spoken-NLU,qa_stock,ALLaM-7B,26,142,165,1.1619718309859155,8.181818181818182,1.0
 MASSIVE-Arabic,spoken-NLU,qa_stock,Qwen2.5-7B,26,142,286,2.0140845070422535,4.72027972027972,1.7333333333333334
@@ -1094,7 +1094,7 @@ MASSIVE-Arabic,spoken-NLU,qa_stock,Mistral-Small-3.1,26,142,250,1.76056338028169
 MASSIVE-Arabic,spoken-NLU,qa_stock,Falcon-H1-7B,26,142,215,1.5140845070422535,6.27906976744186,1.303030303030303
 MASSIVE-Arabic,spoken-NLU,qa_stock,tiktoken-cl100k,26,142,536,3.7746478873239435,2.5186567164179103,3.2484848484848485
 MASSIVE-Arabic,spoken-NLU,qa_stock,tiktoken-o200k,26,142,235,1.6549295774647887,5.74468085106383,1.4242424242424243
-MASSIVE-Arabic,spoken-NLU,recommendation_events,SunAI-64K,43,331,396,1.1963746223564955,8.151515151515152,1.0
+MASSIVE-Arabic,spoken-NLU,recommendation_events,SamAI-Tokenizer,43,331,396,1.1963746223564955,8.151515151515152,1.0
 MASSIVE-Arabic,spoken-NLU,recommendation_events,Fanar-1,43,331,581,1.755287009063444,5.555938037865749,1.4671717171717171
 MASSIVE-Arabic,spoken-NLU,recommendation_events,ALLaM-7B,43,331,383,1.1570996978851964,8.428198433420366,0.9671717171717171
 MASSIVE-Arabic,spoken-NLU,recommendation_events,Qwen2.5-7B,43,331,651,1.9667673716012084,4.95852534562212,1.643939393939394
@@ -1109,7 +1109,7 @@ MASSIVE-Arabic,spoken-NLU,recommendation_events,Mistral-Small-3.1,43,331,531,1.6
 MASSIVE-Arabic,spoken-NLU,recommendation_events,Falcon-H1-7B,43,331,478,1.444108761329305,6.7531380753138075,1.207070707070707
 MASSIVE-Arabic,spoken-NLU,recommendation_events,tiktoken-cl100k,43,331,1187,3.5861027190332324,2.7194608256107835,2.9974747474747474
 MASSIVE-Arabic,spoken-NLU,recommendation_events,tiktoken-o200k,43,331,533,1.6102719033232629,6.056285178236398,1.345959595959596
-MASSIVE-Arabic,spoken-NLU,recommendation_locations,SunAI-64K,31,223,259,1.1614349775784754,7.872586872586872,1.0
+MASSIVE-Arabic,spoken-NLU,recommendation_locations,SamAI-Tokenizer,31,223,259,1.1614349775784754,7.872586872586872,1.0
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,Fanar-1,31,223,395,1.7713004484304933,5.162025316455696,1.5250965250965252
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,ALLaM-7B,31,223,254,1.1390134529147982,8.027559055118111,0.9806949806949807
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,Qwen2.5-7B,31,223,441,1.9775784753363228,4.62358276643991,1.7027027027027026
@@ -1124,7 +1124,7 @@ MASSIVE-Arabic,spoken-NLU,recommendation_locations,Mistral-Small-3.1,31,223,335,
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,Falcon-H1-7B,31,223,337,1.5112107623318385,6.050445103857567,1.3011583011583012
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,tiktoken-cl100k,31,223,772,3.4618834080717487,2.6411917098445596,2.9806949806949805
 MASSIVE-Arabic,spoken-NLU,recommendation_locations,tiktoken-o200k,31,223,361,1.6188340807174888,5.648199445983379,1.393822393822394
-MASSIVE-Arabic,spoken-NLU,recommendation_movies,SunAI-64K,20,118,143,1.2118644067796611,7.6573426573426575,1.0
+MASSIVE-Arabic,spoken-NLU,recommendation_movies,SamAI-Tokenizer,20,118,143,1.2118644067796611,7.6573426573426575,1.0
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,Fanar-1,20,118,197,1.6694915254237288,5.558375634517766,1.3776223776223777
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,ALLaM-7B,20,118,142,1.2033898305084745,7.711267605633803,0.993006993006993
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,Qwen2.5-7B,20,118,222,1.88135593220339,4.9324324324324325,1.5524475524475525
@@ -1139,7 +1139,7 @@ MASSIVE-Arabic,spoken-NLU,recommendation_movies,Mistral-Small-3.1,20,118,174,1.4
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,Falcon-H1-7B,20,118,169,1.4322033898305084,6.479289940828402,1.1818181818181819
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,tiktoken-cl100k,20,118,418,3.542372881355932,2.6196172248803826,2.923076923076923
 MASSIVE-Arabic,spoken-NLU,recommendation_movies,tiktoken-o200k,20,118,188,1.5932203389830508,5.824468085106383,1.3146853146853146
-MASSIVE-Arabic,spoken-NLU,social_post,SunAI-64K,81,546,716,1.3113553113553114,7.209497206703911,1.0
+MASSIVE-Arabic,spoken-NLU,social_post,SamAI-Tokenizer,81,546,716,1.3113553113553114,7.209497206703911,1.0
 MASSIVE-Arabic,spoken-NLU,social_post,Fanar-1,81,546,996,1.8241758241758241,5.182730923694779,1.3910614525139664
 MASSIVE-Arabic,spoken-NLU,social_post,ALLaM-7B,81,546,660,1.2087912087912087,7.821212121212121,0.9217877094972067
 MASSIVE-Arabic,spoken-NLU,social_post,Qwen2.5-7B,81,546,1083,1.9835164835164836,4.766389658356418,1.5125698324022345
@@ -1154,7 +1154,7 @@ MASSIVE-Arabic,spoken-NLU,social_post,Mistral-Small-3.1,81,546,932,1.70695970695
 MASSIVE-Arabic,spoken-NLU,social_post,Falcon-H1-7B,81,546,850,1.5567765567765568,6.0729411764705885,1.187150837988827
 MASSIVE-Arabic,spoken-NLU,social_post,tiktoken-cl100k,81,546,2019,3.697802197802198,2.5567112431896977,2.819832402234637
 MASSIVE-Arabic,spoken-NLU,social_post,tiktoken-o200k,81,546,938,1.7179487179487178,5.50319829424307,1.3100558659217878
-MASSIVE-Arabic,spoken-NLU,social_query,SunAI-64K,25,156,177,1.1346153846153846,8.378531073446327,1.0
+MASSIVE-Arabic,spoken-NLU,social_query,SamAI-Tokenizer,25,156,177,1.1346153846153846,8.378531073446327,1.0
 MASSIVE-Arabic,spoken-NLU,social_query,Fanar-1,25,156,249,1.5961538461538463,5.955823293172691,1.4067796610169492
 MASSIVE-Arabic,spoken-NLU,social_query,ALLaM-7B,25,156,171,1.0961538461538463,8.67251461988304,0.9661016949152542
 MASSIVE-Arabic,spoken-NLU,social_query,Qwen2.5-7B,25,156,269,1.7243589743589745,5.5130111524163565,1.5197740112994351
@@ -1169,7 +1169,7 @@ MASSIVE-Arabic,spoken-NLU,social_query,Mistral-Small-3.1,25,156,241,1.5448717948
 MASSIVE-Arabic,spoken-NLU,social_query,Falcon-H1-7B,25,156,211,1.3525641025641026,7.028436018957346,1.192090395480226
 MASSIVE-Arabic,spoken-NLU,social_query,tiktoken-cl100k,25,156,585,3.75,2.535042735042735,3.305084745762712
 MASSIVE-Arabic,spoken-NLU,social_query,tiktoken-o200k,25,156,239,1.5320512820512822,6.205020920502092,1.3502824858757063
-MASSIVE-Arabic,spoken-NLU,takeaway_order,SunAI-64K,21,80,117,1.4625,6.444444444444445,1.0
+MASSIVE-Arabic,spoken-NLU,takeaway_order,SamAI-Tokenizer,21,80,117,1.4625,6.444444444444445,1.0
 MASSIVE-Arabic,spoken-NLU,takeaway_order,Fanar-1,21,80,161,2.0125,4.683229813664596,1.376068376068376
 MASSIVE-Arabic,spoken-NLU,takeaway_order,ALLaM-7B,21,80,109,1.3625,6.91743119266055,0.9316239316239316
 MASSIVE-Arabic,spoken-NLU,takeaway_order,Qwen2.5-7B,21,80,187,2.3375,4.032085561497326,1.5982905982905984
@@ -1184,7 +1184,7 @@ MASSIVE-Arabic,spoken-NLU,takeaway_order,Mistral-Small-3.1,21,80,156,1.95,4.8333
 MASSIVE-Arabic,spoken-NLU,takeaway_order,Falcon-H1-7B,21,80,154,1.925,4.896103896103896,1.3162393162393162
 MASSIVE-Arabic,spoken-NLU,takeaway_order,tiktoken-cl100k,21,80,306,3.825,2.4640522875816995,2.6153846153846154
 MASSIVE-Arabic,spoken-NLU,takeaway_order,tiktoken-o200k,21,80,159,1.9875,4.7421383647798745,1.358974358974359
-MASSIVE-Arabic,spoken-NLU,takeaway_query,SunAI-64K,30,99,149,1.505050505050505,5.966442953020135,1.0
+MASSIVE-Arabic,spoken-NLU,takeaway_query,SamAI-Tokenizer,30,99,149,1.505050505050505,5.966442953020135,1.0
 MASSIVE-Arabic,spoken-NLU,takeaway_query,Fanar-1,30,99,189,1.9090909090909092,4.703703703703703,1.2684563758389262
 MASSIVE-Arabic,spoken-NLU,takeaway_query,ALLaM-7B,30,99,136,1.3737373737373737,6.536764705882353,0.912751677852349
 MASSIVE-Arabic,spoken-NLU,takeaway_query,Qwen2.5-7B,30,99,213,2.1515151515151514,4.173708920187793,1.429530201342282
@@ -1199,7 +1199,7 @@ MASSIVE-Arabic,spoken-NLU,takeaway_query,Mistral-Small-3.1,30,99,176,1.777777777
 MASSIVE-Arabic,spoken-NLU,takeaway_query,Falcon-H1-7B,30,99,176,1.7777777777777777,5.051136363636363,1.1812080536912752
 MASSIVE-Arabic,spoken-NLU,takeaway_query,tiktoken-cl100k,30,99,368,3.717171717171717,2.4157608695652173,2.469798657718121
 MASSIVE-Arabic,spoken-NLU,takeaway_query,tiktoken-o200k,30,99,180,1.8181818181818181,4.938888888888889,1.2080536912751678
-MASSIVE-Arabic,spoken-NLU,transport_query,SunAI-64K,43,138,169,1.2246376811594204,8.011834319526628,1.0
+MASSIVE-Arabic,spoken-NLU,transport_query,SamAI-Tokenizer,43,138,169,1.2246376811594204,8.011834319526628,1.0
 MASSIVE-Arabic,spoken-NLU,transport_query,Fanar-1,43,138,255,1.8478260869565217,5.309803921568627,1.5088757396449703
 MASSIVE-Arabic,spoken-NLU,transport_query,ALLaM-7B,43,138,154,1.1159420289855073,8.792207792207792,0.9112426035502958
 MASSIVE-Arabic,spoken-NLU,transport_query,Qwen2.5-7B,43,138,265,1.9202898550724639,5.109433962264151,1.5680473372781065
@@ -1214,7 +1214,7 @@ MASSIVE-Arabic,spoken-NLU,transport_query,Mistral-Small-3.1,43,138,240,1.7391304
 MASSIVE-Arabic,spoken-NLU,transport_query,Falcon-H1-7B,43,138,216,1.565217391304348,6.268518518518518,1.2781065088757397
 MASSIVE-Arabic,spoken-NLU,transport_query,tiktoken-cl100k,43,138,524,3.7971014492753623,2.5839694656488548,3.100591715976331
 MASSIVE-Arabic,spoken-NLU,transport_query,tiktoken-o200k,43,138,239,1.7318840579710144,5.665271966527197,1.4142011834319526
-MASSIVE-Arabic,spoken-NLU,transport_taxi,SunAI-64K,21,63,95,1.507936507936508,6.336842105263158,1.0
+MASSIVE-Arabic,spoken-NLU,transport_taxi,SamAI-Tokenizer,21,63,95,1.507936507936508,6.336842105263158,1.0
 MASSIVE-Arabic,spoken-NLU,transport_taxi,Fanar-1,21,63,128,2.0317460317460316,4.703125,1.3473684210526315
 MASSIVE-Arabic,spoken-NLU,transport_taxi,ALLaM-7B,21,63,80,1.2698412698412698,7.525,0.8421052631578947
 MASSIVE-Arabic,spoken-NLU,transport_taxi,Qwen2.5-7B,21,63,137,2.1746031746031744,4.394160583941606,1.4421052631578948
@@ -1229,7 +1229,7 @@ MASSIVE-Arabic,spoken-NLU,transport_taxi,Mistral-Small-3.1,21,63,118,1.873015873
 MASSIVE-Arabic,spoken-NLU,transport_taxi,Falcon-H1-7B,21,63,117,1.8571428571428572,5.145299145299146,1.231578947368421
 MASSIVE-Arabic,spoken-NLU,transport_taxi,tiktoken-cl100k,21,63,239,3.7936507936507935,2.518828451882845,2.5157894736842104
 MASSIVE-Arabic,spoken-NLU,transport_taxi,tiktoken-o200k,21,63,128,2.0317460317460316,4.703125,1.3473684210526315
-MASSIVE-Arabic,spoken-NLU,transport_ticket,SunAI-64K,32,122,141,1.1557377049180328,8.73758865248227,1.0
+MASSIVE-Arabic,spoken-NLU,transport_ticket,SamAI-Tokenizer,32,122,141,1.1557377049180328,8.73758865248227,1.0
 MASSIVE-Arabic,spoken-NLU,transport_ticket,Fanar-1,32,122,247,2.0245901639344264,4.987854251012146,1.75177304964539
 MASSIVE-Arabic,spoken-NLU,transport_ticket,ALLaM-7B,32,122,132,1.0819672131147542,9.333333333333334,0.9361702127659575
 MASSIVE-Arabic,spoken-NLU,transport_ticket,Qwen2.5-7B,32,122,253,2.0737704918032787,4.869565217391305,1.7943262411347518
@@ -1244,7 +1244,7 @@ MASSIVE-Arabic,spoken-NLU,transport_ticket,Mistral-Small-3.1,32,122,226,1.852459
 MASSIVE-Arabic,spoken-NLU,transport_ticket,Falcon-H1-7B,32,122,195,1.598360655737705,6.317948717948718,1.3829787234042554
 MASSIVE-Arabic,spoken-NLU,transport_ticket,tiktoken-cl100k,32,122,486,3.9836065573770494,2.5349794238683128,3.4468085106382977
 MASSIVE-Arabic,spoken-NLU,transport_ticket,tiktoken-o200k,32,122,229,1.8770491803278688,5.379912663755459,1.624113475177305
-MASSIVE-Arabic,spoken-NLU,transport_traffic,SunAI-64K,13,29,34,1.1724137931034482,8.058823529411764,1.0
+MASSIVE-Arabic,spoken-NLU,transport_traffic,SamAI-Tokenizer,13,29,34,1.1724137931034482,8.058823529411764,1.0
 MASSIVE-Arabic,spoken-NLU,transport_traffic,Fanar-1,13,29,53,1.8275862068965518,5.169811320754717,1.5588235294117647
 MASSIVE-Arabic,spoken-NLU,transport_traffic,ALLaM-7B,13,29,31,1.0689655172413792,8.838709677419354,0.9117647058823529
 MASSIVE-Arabic,spoken-NLU,transport_traffic,Qwen2.5-7B,13,29,56,1.9310344827586208,4.892857142857143,1.6470588235294117
@@ -1259,7 +1259,7 @@ MASSIVE-Arabic,spoken-NLU,transport_traffic,Mistral-Small-3.1,13,29,49,1.6896551
 MASSIVE-Arabic,spoken-NLU,transport_traffic,Falcon-H1-7B,13,29,45,1.5517241379310345,6.088888888888889,1.3235294117647058
 MASSIVE-Arabic,spoken-NLU,transport_traffic,tiktoken-cl100k,13,29,108,3.7241379310344827,2.537037037037037,3.176470588235294
 MASSIVE-Arabic,spoken-NLU,transport_traffic,tiktoken-o200k,13,29,50,1.7241379310344827,5.48,1.4705882352941178
-MASSIVE-Arabic,spoken-NLU,weather_query,SunAI-64K,153,854,978,1.14519906323185,7.782208588957055,1.0
+MASSIVE-Arabic,spoken-NLU,weather_query,SamAI-Tokenizer,153,854,978,1.14519906323185,7.782208588957055,1.0
 MASSIVE-Arabic,spoken-NLU,weather_query,Fanar-1,153,854,1448,1.6955503512880563,5.2562154696132595,1.4805725971370143
 MASSIVE-Arabic,spoken-NLU,weather_query,ALLaM-7B,153,854,944,1.1053864168618266,8.0625,0.9652351738241309
 MASSIVE-Arabic,spoken-NLU,weather_query,Qwen2.5-7B,153,854,1444,1.6908665105386418,5.270775623268698,1.4764826175869121
@@ -1274,7 +1274,7 @@ MASSIVE-Arabic,spoken-NLU,weather_query,Mistral-Small-3.1,153,854,1257,1.4718969
 MASSIVE-Arabic,spoken-NLU,weather_query,Falcon-H1-7B,153,854,1143,1.338407494145199,6.6587926509186355,1.1687116564417177
 MASSIVE-Arabic,spoken-NLU,weather_query,tiktoken-cl100k,153,854,2990,3.5011709601873537,2.545484949832776,3.0572597137014315
 MASSIVE-Arabic,spoken-NLU,weather_query,tiktoken-o200k,153,854,1281,1.5,5.941451990632318,1.3098159509202454
-MLQA-Arabic,QA,passages,SunAI-64K,5085,594082,859911,1.4474617982029416,7.3901892172562045,1.0
+MLQA-Arabic,QA,passages,SamAI-Tokenizer,5085,594082,859911,1.4474617982029416,7.3901892172562045,1.0
 MLQA-Arabic,QA,passages,Fanar-1,5085,594082,1265617,2.1303742580990503,5.0211912450607095,1.4717999886034718
 MLQA-Arabic,QA,passages,ALLaM-7B,5085,594082,921314,1.550819583828495,6.897653785788559,1.0714062269234839
 MLQA-Arabic,QA,passages,Qwen2.5-7B,5085,594082,1391740,2.3426732336613463,4.56615819046661,1.618469818388182
@@ -1289,7 +1289,7 @@ MLQA-Arabic,QA,passages,Mistral-Small-3.1,5085,594082,1088662,1.832511336818823,
 MLQA-Arabic,QA,passages,Falcon-H1-7B,5085,594082,1089060,1.833181278005393,5.835220281710833,1.266479903152768
 MLQA-Arabic,QA,passages,tiktoken-cl100k,5085,594082,2458220,4.137846290579415,2.5851652821960607,2.858691190134793
 MLQA-Arabic,QA,passages,tiktoken-o200k,5085,594082,1153289,1.9412959827094576,5.510245046991691,1.3411725166906807
-MLQA-Arabic,QA,questions,SunAI-64K,5848,44626,59616,1.3359028369112176,7.399372651637145,1.0
+MLQA-Arabic,QA,questions,SamAI-Tokenizer,5848,44626,59616,1.3359028369112176,7.399372651637145,1.0
 MLQA-Arabic,QA,questions,Fanar-1,5848,44626,84325,1.8895935105095685,5.23120071153276,1.4144692699946324
 MLQA-Arabic,QA,questions,ALLaM-7B,5848,44626,61678,1.3821090843902657,7.151999092058757,1.0345880300590446
 MLQA-Arabic,QA,questions,Qwen2.5-7B,5848,44626,92872,2.0811186303948372,4.749773882332673,1.557836822329576

results/arabic_4bench_macro_average.csv

@@ -1,6 +1,6 @@
 rank,tokenizer,benchmarks,mean_tokens_per_word,mean_characters_per_token,mean_bytes_per_token,mean_token_ratio_vs_yours,mean_extra_tokens_vs_yours_pct
 1,ALLaM-7B,4,1.502087831475563,3.8909346754141927,7.034872305543145,0.9513816695259771,-4.861833047402281
-2,SunAI-64K,4,1.6206788856631889,3.72984537725556,6.739888959283761,1.0,0.0
+2,SamAI-Tokenizer,4,1.6206788856631889,3.72984537725556,6.739888959283761,1.0,0.0
 3,Falcon-H1-7B,4,1.8879948366394974,3.0986638596160905,5.601525945236872,1.1929511423268218,19.295114232682163
 4,Mistral-Small-3.1,4,1.8929229164641286,3.0610566424373293,5.534005895771695,1.2071316671047543,20.713166710475434
 5,Qwen3.8-Flash-Next,4,1.9038564910552505,3.0317074563171937,5.481727181321753,1.2194430010126256,21.94430010126255

results/arabic_4bench_pooled_overall.csv

@@ -1,6 +1,6 @@
 benchmark,tokenizer,documents,tokens,tokens_per_word,characters_per_token,bytes_per_token,token_ratio_vs_yours,extra_tokens_vs_yours_pct
 ALL-POOLED,ALLaM-7B,38082,2252701,1.6064524778521883,3.6839460718488604,6.659080810103072,0.939031547811095,-6.0968452188905005
-ALL-POOLED,SunAI-64K,38082,2398962,1.7107545338565753,3.459341581900839,6.253086960110248,1.0,0.0
+ALL-POOLED,SamAI-Tokenizer,38082,2398962,1.7107545338565753,3.459341581900839,6.253086960110248,1.0,0.0
 ALL-POOLED,Mistral-Small-3.1,38082,2761120,1.969017666191489,3.0056024366923566,5.432910558034421,1.1509644587950956,15.096445879509556
 ALL-POOLED,Falcon-H1-7B,38082,2783770,1.9851698979449939,2.98114750859446,5.388705963495547,1.1604060422799527,16.04060422799527
 ALL-POOLED,Qwen3.8-Flash-Next,38082,2793312,1.991974515843093,2.970963859389857,5.37029805478228,1.164383595905229,16.43835959052291

results/arabicmmlu_by_subject.csv

@@ -1,5 +1,5 @@
 group,subject,tokenizer,documents,words,tokens,tokens_per_word,bytes_per_token,token_ratio_vs_yours
-Humanities,History,SunAI-64K,1065,22191,33607,1.5144427921229326,7.207873359716726,1.0
+Humanities,History,SamAI-Tokenizer,1065,22191,33607,1.5144427921229326,7.207873359716726,1.0
 Humanities,History,Fanar-1,1065,22191,48439,2.1828218647199313,5.000825780879044,1.4413366262980927
 Humanities,History,ALLaM-7B,1065,22191,36070,1.6254337343968275,6.715691710562795,1.0732883030321063
 Humanities,History,Qwen2.5-7B,1065,22191,51053,2.3006173674011987,4.744775037705914,1.5191180408843397
@@ -13,7 +13,7 @@ Humanities,History,Mistral-Small-3.1,1065,22191,42654,1.9221305934838449,5.67906
 Humanities,History,Falcon-H1-7B,1065,22191,42137,1.8988328601685367,5.748748131096186,1.2538161692504537
 Humanities,History,tiktoken-cl100k,1065,22191,92850,4.184128700824659,2.608885298869144,2.762817270211563
 Humanities,History,tiktoken-o200k,1065,22191,43976,1.98170429453382,5.5083454611606335,1.308536911952867
-Humanities,Islamic Studies,SunAI-64K,2210,39174,61563,1.5715270332363303,6.147994737098582,1.0
+Humanities,Islamic Studies,SamAI-Tokenizer,2210,39174,61563,1.5715270332363303,6.147994737098582,1.0
 Humanities,Islamic Studies,Fanar-1,2210,39174,81872,2.0899576249553276,4.622935802227868,1.3298897064795412
 Humanities,Islamic Studies,ALLaM-7B,2210,39174,62341,1.5913871445346404,6.071269309122408,1.012637460812501
 Humanities,Islamic Studies,Qwen2.5-7B,2210,39174,85998,2.195282585388268,4.401137235749669,1.3969104819453244
@@ -27,7 +27,7 @@ Humanities,Islamic Studies,Mistral-Small-3.1,2210,39174,73369,1.8729003931178843
 Humanities,Islamic Studies,Falcon-H1-7B,2210,39174,75589,1.9295706335834992,5.007196814351294,1.2278316521287136
 Humanities,Islamic Studies,tiktoken-cl100k,2210,39174,153064,3.9072854444274263,2.472749960800711,2.4862985884378603
 Humanities,Islamic Studies,tiktoken-o200k,2210,39174,76229,1.945908000204217,4.9651576171798135,1.2382275067816708
-Humanities,Law,SunAI-64K,314,11281,15675,1.389504476553497,7.622838915470495,1.0
+Humanities,Law,SamAI-Tokenizer,314,11281,15675,1.389504476553497,7.622838915470495,1.0
 Humanities,Law,Fanar-1,314,11281,23510,2.0840351032709865,5.082433007230965,1.4998405103668262
 Humanities,Law,ALLaM-7B,314,11281,15815,1.4019147238719971,7.5553588365475814,1.0089314194577352
 Humanities,Law,Qwen2.5-7B,314,11281,24968,2.213278964630795,4.785645626401794,1.5928548644338119
@@ -41,7 +41,7 @@ Humanities,Law,Mistral-Small-3.1,314,11281,20248,1.7948763407499335,5.9012248123
 Humanities,Law,Falcon-H1-7B,314,11281,19290,1.709954791241911,6.194297563504406,1.230622009569378
 Humanities,Law,tiktoken-cl100k,314,11281,47266,4.189876783973052,2.5279905217280922,3.0153748006379586
 Humanities,Law,tiktoken-o200k,314,11281,21664,1.920397127914192,5.515509601181684,1.38207336523126
-Humanities,Philosophy,SunAI-64K,39,1407,2160,1.535181236673774,6.564814814814815,1.0
+Humanities,Philosophy,SamAI-Tokenizer,39,1407,2160,1.535181236673774,6.564814814814815,1.0
 Humanities,Philosophy,Fanar-1,39,1407,2989,2.1243781094527363,4.744061559049849,1.3837962962962962
 Humanities,Philosophy,ALLaM-7B,39,1407,2229,1.584221748400853,6.36159712875729,1.0319444444444446
 Humanities,Philosophy,Qwen2.5-7B,39,1407,3076,2.1862117981520965,4.609882964889467,1.424074074074074
@@ -55,7 +55,7 @@ Humanities,Philosophy,Mistral-Small-3.1,39,1407,2579,1.8329779673063256,5.498255
 Humanities,Philosophy,Falcon-H1-7B,39,1407,2839,2.017768301350391,4.9947164494540335,1.3143518518518518
 Humanities,Philosophy,tiktoken-cl100k,39,1407,5530,3.9303482587064678,2.5641952983725136,2.560185185185185
 Humanities,Philosophy,tiktoken-o200k,39,1407,2701,1.919687277896233,5.249907441688263,1.2504629629629629
-Language,Arabic Language,SunAI-64K,669,17268,32934,1.9072272411396802,5.384982085382887,1.0
+Language,Arabic Language,SamAI-Tokenizer,669,17268,32934,1.9072272411396802,5.384982085382887,1.0
 Language,Arabic Language,Fanar-1,669,17268,41489,2.402652304841325,4.274602906794572,1.2597619481387017
 Language,Arabic Language,ALLaM-7B,669,17268,31908,1.847810979847116,5.558135890685722,0.968846784478047
 Language,Arabic Language,Qwen2.5-7B,669,17268,43938,2.5444753300903407,4.036346670308162,1.3341227910366187
@@ -69,7 +69,7 @@ Language,Arabic Language,Mistral-Small-3.1,669,17268,37280,2.158906648135279,4.7
 Language,Arabic Language,Falcon-H1-7B,669,17268,38668,2.239286541579801,4.586453915382228,1.1741057873322402
 Language,Arabic Language,tiktoken-cl100k,669,17268,72894,4.221334259902711,2.43297116360743,2.213335762433959
 Language,Arabic Language,tiktoken-o200k,669,17268,38810,2.2475098447996293,4.569672764751353,1.178417440942491
-Language,Arabic Language (General),SunAI-64K,612,130856,563958,4.309760347251941,3.265697090918118,1.0
+Language,Arabic Language (General),SamAI-Tokenizer,612,130856,563958,4.309760347251941,3.265697090918118,1.0
 Language,Arabic Language (General),Fanar-1,612,130856,544644,4.162162988323042,3.3815042486468223,0.9657527688232103
 Language,Arabic Language (General),ALLaM-7B,612,130856,340840,2.6046952375129915,5.403462034972421,0.6043712475042468
 Language,Arabic Language (General),Qwen2.5-7B,612,130856,638412,4.878736932200281,2.8848392574074424,1.132020469609439
@@ -83,7 +83,7 @@ Language,Arabic Language (General),Mistral-Small-3.1,612,130856,500913,3.8279712
 Language,Arabic Language (General),Falcon-H1-7B,612,130856,485719,3.7118588371950847,3.7917314331949132,0.86126803769075
 Language,Arabic Language (General),tiktoken-cl100k,612,130856,800767,6.119451916610625,2.2999399325896297,1.419905383024977
 Language,Arabic Language (General),tiktoken-o200k,612,130856,507338,3.877070978785841,3.6301558329949657,0.8996024526649148
-Language,Arabic Language (Grammar),SunAI-64K,365,9238,18334,1.9846287075124487,5.132158830587978,1.0
+Language,Arabic Language (Grammar),SamAI-Tokenizer,365,9238,18334,1.9846287075124487,5.132158830587978,1.0
 Language,Arabic Language (Grammar),Fanar-1,365,9238,23311,2.523381684347261,4.03642057397795,1.2714628558961492
 Language,Arabic Language (Grammar),ALLaM-7B,365,9238,16880,1.8272353323230137,5.574229857819906,0.9206937929529835
 Language,Arabic Language (Grammar),Qwen2.5-7B,365,9238,24349,2.6357436674604893,3.8643476118115734,1.3280789789462202
@@ -97,7 +97,7 @@ Language,Arabic Language (Grammar),Mistral-Small-3.1,365,9238,20445,2.2131413725
 Language,Arabic Language (Grammar),Falcon-H1-7B,365,9238,21652,2.343797358735657,4.345695547755404,1.1809752372640996
 Language,Arabic Language (Grammar),tiktoken-cl100k,365,9238,39172,4.240311755791296,2.402047380782191,2.136576851750845
 Language,Arabic Language (Grammar),tiktoken-o200k,365,9238,21320,2.307858843905607,4.413367729831145,1.1628668048434603
-Other,Driving Test,SunAI-64K,1211,29647,44828,1.5120585556717374,6.785602748282323,1.0
+Other,Driving Test,SamAI-Tokenizer,1211,29647,44828,1.5120585556717374,6.785602748282323,1.0
 Other,Driving Test,Fanar-1,1211,29647,63874,2.15448443350086,4.762266336850675,1.4248683858302846
 Other,Driving Test,ALLaM-7B,1211,29647,47040,1.5866698148210612,6.466517857142857,1.0493441599000624
 Other,Driving Test,Qwen2.5-7B,1211,29647,68843,2.3220899247815967,4.4185320221373265,1.5357142857142858
@@ -111,7 +111,7 @@ Other,Driving Test,Mistral-Small-3.1,1211,29647,55718,1.8793807130569704,5.45936
 Other,Driving Test,Falcon-H1-7B,1211,29647,54844,1.849900495834317,5.546367879804537,1.223431783706612
 Other,Driving Test,tiktoken-cl100k,1211,29647,119858,4.042837386582116,2.5378781558177175,2.6737307040242704
 Other,Driving Test,tiktoken-o200k,1211,29647,57716,1.9467737039160793,5.270375632406958,1.2874988846256803
-Other,General Knowledge,SunAI-64K,1198,19003,32538,1.7122559595853286,6.0611592599422215,1.0
+Other,General Knowledge,SamAI-Tokenizer,1198,19003,32538,1.7122559595853286,6.0611592599422215,1.0
 Other,General Knowledge,Fanar-1,1198,19003,44258,2.32900068410251,4.456098332504858,1.3601942344335853
 Other,General Knowledge,ALLaM-7B,1198,19003,34096,1.7942430142609063,5.784197559831065,1.0478824758743623
 Other,General Knowledge,Qwen2.5-7B,1198,19003,46807,2.463137399357996,4.213429615228492,1.4385334070932447
@@ -125,7 +125,7 @@ Other,General Knowledge,Mistral-Small-3.1,1198,19003,39568,2.0821975477556176,4.
 Other,General Knowledge,Falcon-H1-7B,1198,19003,40229,2.1169815292322265,4.902383852444753,1.236369783022927
 Other,General Knowledge,tiktoken-cl100k,1198,19003,80452,4.233647318844393,2.4513747327599065,2.4725551662671337
 Other,General Knowledge,tiktoken-o200k,1198,19003,40827,2.1484502446982057,4.830577803904279,1.254748294302047
-Other,Management,SunAI-64K,75,1686,2805,1.6637010676156583,6.080213903743315,1.0
+Other,Management,SamAI-Tokenizer,75,1686,2805,1.6637010676156583,6.080213903743315,1.0
 Other,Management,Fanar-1,75,1686,3825,2.2686832740213525,4.458823529411765,1.3636363636363635
 Other,Management,ALLaM-7B,75,1686,2906,1.723606168446026,5.868891947694426,1.036007130124777
 Other,Management,Qwen2.5-7B,75,1686,3811,2.260379596678529,4.475203358698504,1.3586452762923351
@@ -139,7 +139,7 @@ Other,Management,Mistral-Small-3.1,75,1686,3251,1.9282325029655991,5.24607812980
 Other,Management,Falcon-H1-7B,75,1686,3347,1.9851720047449586,5.0956080071706005,1.1932263814616755
 Other,Management,tiktoken-cl100k,75,1686,6753,4.005338078291815,2.525544202576633,2.407486631016043
 Other,Management,tiktoken-o200k,75,1686,3395,2.013641755634638,5.023564064801178,1.2103386809269163
-STEM,Biology,SunAI-64K,1409,29117,47477,1.630559466978054,6.448048528761295,1.0
+STEM,Biology,SamAI-Tokenizer,1409,29117,47477,1.630559466978054,6.448048528761295,1.0
 STEM,Biology,Fanar-1,1409,29117,67911,2.332348799670296,4.507870595338016,1.4303978768666934
 STEM,Biology,ALLaM-7B,1409,29117,51255,1.7603118453137343,5.972763632816311,1.0795753733386693
 STEM,Biology,Qwen2.5-7B,1409,29117,72478,2.489198749871209,4.223819641822346,1.5265918234092297
@@ -153,7 +153,7 @@ STEM,Biology,Mistral-Small-3.1,1409,29117,60698,2.084624102757839,5.043559919601
 STEM,Biology,Falcon-H1-7B,1409,29117,61629,2.1165985506748637,4.967369257979198,1.2980811761484508
 STEM,Biology,tiktoken-cl100k,1409,29117,121687,4.179242367002095,2.51574942269922,2.5630726457021296
 STEM,Biology,tiktoken-o200k,1409,29117,63269,2.1729230346532953,4.838609745689042,1.3326242180424206
-STEM,Computer Science,SunAI-64K,542,9776,14540,1.487315875613748,7.119463548830812,1.0
+STEM,Computer Science,SamAI-Tokenizer,542,9776,14540,1.487315875613748,7.119463548830812,1.0
 STEM,Computer Science,Fanar-1,542,9776,21367,2.1856587561374794,4.844713811016989,1.469532324621733
 STEM,Computer Science,ALLaM-7B,542,9776,15718,1.6078150572831424,6.585888789922382,1.0810178817056395
 STEM,Computer Science,Qwen2.5-7B,542,9776,22331,2.2842675941080195,4.635573865926291,1.5358321870701512
@@ -167,7 +167,7 @@ STEM,Computer Science,Mistral-Small-3.1,542,9776,18821,1.9252250409165304,5.5000
 STEM,Computer Science,Falcon-H1-7B,542,9776,19322,1.9764729950900164,5.357468170996791,1.3288858321870702
 STEM,Computer Science,tiktoken-cl100k,542,9776,40046,4.096358428805237,2.584952304849423,2.7541953232462175
 STEM,Computer Science,tiktoken-o200k,542,9776,19491,1.9937602291325696,5.311015340413524,1.34050894085282
-STEM,Math,SunAI-64K,409,4585,7744,1.6889858233369683,4.202221074380165,1.0
+STEM,Math,SamAI-Tokenizer,409,4585,7744,1.6889858233369683,4.202221074380165,1.0
 STEM,Math,Fanar-1,409,4585,11268,2.4575790621592146,2.8880014199503017,1.4550619834710743
 STEM,Math,ALLaM-7B,409,4585,9977,2.176008724100327,3.2617019144031274,1.2883522727272727
 STEM,Math,Qwen2.5-7B,409,4585,11670,2.5452562704471102,2.788517566409597,1.5069731404958677
@@ -181,7 +181,7 @@ STEM,Math,Mistral-Small-3.1,409,4585,10657,2.324318429661941,3.0535798066998217,
 STEM,Math,Falcon-H1-7B,409,4585,11069,2.414176663031625,2.939922305537989,1.4293646694214877
 STEM,Math,tiktoken-cl100k,409,4585,14486,3.159432933478735,2.2464448432969766,1.8706095041322315
 STEM,Math,tiktoken-o200k,409,4585,9200,2.0065430752453652,3.5371739130434783,1.18801652892562
-STEM,Natural Science,SunAI-64K,578,9714,17306,1.7815523985999588,6.16751415693979,1.0
+STEM,Natural Science,SamAI-Tokenizer,578,9714,17306,1.7815523985999588,6.16751415693979,1.0
 STEM,Natural Science,Fanar-1,578,9714,23515,2.4207329627341982,4.539017648309589,1.3587773026695944
 STEM,Natural Science,ALLaM-7B,578,9714,17754,1.8276714021000617,6.011884645713642,1.0258869756153934
 STEM,Natural Science,Qwen2.5-7B,578,9714,25268,2.601194152769199,4.224117460820009,1.460071651450364
@@ -195,7 +195,7 @@ STEM,Natural Science,Mistral-Small-3.1,578,9714,20641,2.1248713197446985,5.17101
 STEM,Natural Science,Falcon-H1-7B,578,9714,21462,2.209388511426807,4.973208461466778,1.240147925574945
 STEM,Natural Science,tiktoken-cl100k,578,9714,42341,4.358760551780935,2.5208426820339622,2.446608112793251
 STEM,Natural Science,tiktoken-o200k,578,9714,21630,2.2266831377393452,4.934581599630143,1.2498555414307178
-STEM,Physics,SunAI-64K,255,7398,11407,1.5419032170856988,5.738406241781362,1.0
+STEM,Physics,SamAI-Tokenizer,255,7398,11407,1.5419032170856988,5.738406241781362,1.0
 STEM,Physics,Fanar-1,255,7398,16104,2.1768045417680453,4.064704421261799,1.411764705882353
 STEM,Physics,ALLaM-7B,255,7398,12596,1.7026223303595567,5.196729120355668,1.1042342421320241
 STEM,Physics,Qwen2.5-7B,255,7398,17613,2.380778588807786,3.7164594333730765,1.5440518979573945
@@ -209,7 +209,7 @@ STEM,Physics,Mistral-Small-3.1,255,7398,14345,1.9390375777237092,4.5631230393865
 STEM,Physics,Falcon-H1-7B,255,7398,14855,2.0079751284130847,4.406462470548637,1.302270535636013
 STEM,Physics,tiktoken-cl100k,255,7398,27688,3.742633144092998,2.3641288644900316,2.4272814938195846
 STEM,Physics,tiktoken-o200k,255,7398,14807,2.0014868883482024,4.420746944012967,1.2980625931445604
-Social Science,Accounting,SunAI-64K,74,2202,4070,1.8483197093551318,5.774447174447174,1.0
+Social Science,Accounting,SamAI-Tokenizer,74,2202,4070,1.8483197093551318,5.774447174447174,1.0
 Social Science,Accounting,Fanar-1,74,2202,5447,2.4736603088101727,4.314668624931155,1.3383292383292382
 Social Science,Accounting,ALLaM-7B,74,2202,4242,1.9264305177111716,5.5403111739745405,1.0422604422604422
 Social Science,Accounting,Qwen2.5-7B,74,2202,5808,2.637602179836512,4.046487603305785,1.427027027027027
@@ -223,7 +223,7 @@ Social Science,Accounting,Mistral-Small-3.1,74,2202,4963,2.2538601271571297,4.73
 Social Science,Accounting,Falcon-H1-7B,74,2202,5037,2.287465940054496,4.665872543180464,1.2375921375921375
 Social Science,Accounting,tiktoken-cl100k,74,2202,9647,4.381017257039056,2.436197781693791,2.3702702702702703
 Social Science,Accounting,tiktoken-o200k,74,2202,5048,2.292461398728429,4.655705229793978,1.2402948402948404
-Social Science,Civics,SunAI-64K,323,7045,10769,1.5286018452803407,7.218590398365679,1.0
+Social Science,Civics,SamAI-Tokenizer,323,7045,10769,1.5286018452803407,7.218590398365679,1.0
 Social Science,Civics,Fanar-1,323,7045,15673,2.2246983676366217,4.9599310916863395,1.455381186739716
 Social Science,Civics,ALLaM-7B,323,7045,11696,1.6601845280340668,6.6464603283173735,1.0860804160089146
 Social Science,Civics,Qwen2.5-7B,323,7045,16640,2.361958836053939,4.6716947115384615,1.5451759680564583
@@ -237,7 +237,7 @@ Social Science,Civics,Mistral-Small-3.1,323,7045,13880,1.9701916252661462,5.6006
 Social Science,Civics,Falcon-H1-7B,323,7045,13354,1.895528743789922,5.8212520593080725,1.2400408580183861
 Social Science,Civics,tiktoken-cl100k,323,7045,30289,4.299361249112846,2.5665092938030307,2.812610270220076
 Social Science,Civics,tiktoken-o200k,323,7045,14008,1.988360539389638,5.549471730439748,1.3007707308013743
-Social Science,Economics,SunAI-64K,584,15822,24265,1.5336240677537605,7.296723676076653,1.0
+Social Science,Economics,SamAI-Tokenizer,584,15822,24265,1.5336240677537605,7.296723676076653,1.0
 Social Science,Economics,Fanar-1,584,15822,35489,2.2430160535962584,4.989010679365437,1.462559241706161
 Social Science,Economics,ALLaM-7B,584,15822,25510,1.6123119706737454,6.9406115248922,1.0513084689882546
 Social Science,Economics,Qwen2.5-7B,584,15822,37455,2.3672734167614715,4.727139233747163,1.5435812899237584
@@ -251,7 +251,7 @@ Social Science,Economics,Mistral-Small-3.1,584,15822,31225,1.9735178864871699,5.
 Social Science,Economics,Falcon-H1-7B,584,15822,29581,1.8696119327518645,5.985429836719516,1.2190809808365959
 Social Science,Economics,tiktoken-cl100k,584,15822,67099,4.2408671470104915,2.6387129465416774,2.7652586029260253
 Social Science,Economics,tiktoken-o200k,584,15822,31515,1.9918467956010617,5.618118356338251,1.2987842571605193
-Social Science,Geography,SunAI-64K,1367,27069,43937,1.6231482507665596,6.795479891663063,1.0
+Social Science,Geography,SamAI-Tokenizer,1367,27069,43937,1.6231482507665596,6.795479891663063,1.0
 Social Science,Geography,Fanar-1,1367,27069,62704,2.3164505522922902,4.76162605256443,1.4271343059380477
 Social Science,Geography,ALLaM-7B,1367,27069,46879,1.7318334626325316,6.369013844151966,1.0669595102077976
 Social Science,Geography,Qwen2.5-7B,1367,27069,67231,2.4836898296944843,4.441001918757716,1.5301681953706443
@@ -265,7 +265,7 @@ Social Science,Geography,Mistral-Small-3.1,1367,27069,54771,2.0233846835863902,5
 Social Science,Geography,Falcon-H1-7B,1367,27069,55178,2.0384203332225055,5.411087752365073,1.2558435942372033
 Social Science,Geography,tiktoken-cl100k,1367,27069,115984,4.284753777383723,2.5742602427921093,2.639779684548331
 Social Science,Geography,tiktoken-o200k,1367,27069,56698,2.094573127932321,5.266023492892166,1.290438582515875
-Social Science,Political Science,SunAI-64K,210,3858,10023,2.5979782270606533,4.5571186271575375,1.0
+Social Science,Political Science,SamAI-Tokenizer,210,3858,10023,2.5979782270606533,4.5571186271575375,1.0
 Social Science,Political Science,Fanar-1,210,3858,12004,3.111456713322965,3.8050649783405532,1.1976454155442482
 Social Science,Political Science,ALLaM-7B,210,3858,12164,3.152928978745464,3.7550147977638932,1.2136086999900229
 Social Science,Political Science,Qwen2.5-7B,210,3858,12639,3.276049766718507,3.6138935042329297,1.2609997006884166
@@ -279,7 +279,7 @@ Social Science,Political Science,Mistral-Small-3.1,210,3858,14998,3.887506480041
 Social Science,Political Science,Falcon-H1-7B,210,3858,12929,3.351218247796786,3.5328331657514114,1.2899331537463834
 Social Science,Political Science,tiktoken-cl100k,210,3858,22315,5.784085018144116,2.046874299798342,2.226379327546643
 Social Science,Political Science,tiktoken-o200k,210,3858,13506,3.5007776049766717,3.3819043388123795,1.3475007482789585
-Social Science,Social Science,SunAI-64K,946,12617,20441,1.6201157168899105,6.404823638765227,1.0
+Social Science,Social Science,SamAI-Tokenizer,946,12617,20441,1.6201157168899105,6.404823638765227,1.0
 Social Science,Social Science,Fanar-1,946,12617,28272,2.2407862407862407,4.6307654216185625,1.383102587936011
 Social Science,Social Science,ALLaM-7B,946,12617,21541,1.7072996750416105,6.0777586927255,1.0538134142165256
 Social Science,Social Science,Qwen2.5-7B,946,12617,29450,2.3341523341523343,4.44553480475382,1.4407318624333447

results/arabicmmlu_summary.csv

@@ -2,7 +2,7 @@ tokenizer,documents,tokens,tokens_per_word,characters_per_token,bytes_per_token,
 ALLaM-7B,14455,819457,1.9940358288275573,3.17696962744842,5.801229350655373,0.8030892382355218,-19.69107617644782
 Qwen3.5-4B,14455,985479,2.3980275164616964,2.6417508643005077,4.82390593812755,0.9657951294663464,-3.420487053365362
 Qwen3.8-Flash-Next,14455,985479,2.3980275164616964,2.6417508643005077,4.82390593812755,0.9657951294663464,-3.420487053365362
-SunAI-64K,14455,1020381,2.4829567299503106,2.5513901180049414,4.658904860047374,1.0,0.0
+SamAI-Tokenizer,14455,1020381,2.4829567299503106,2.5513901180049414,4.658904860047374,1.0,0.0
 Falcon-H1-7B,14455,1054020,2.564812606763774,2.4699626193051367,4.510216124931215,1.0329670975841376,3.296709758413763
 Mistral-Small-3.1,14455,1065947,2.593835319768149,2.442325931777096,4.459750813126732,1.0446558687392258,4.465586873922578
 tiktoken-o200k,14455,1088838,2.649537417813186,2.3909801090704037,4.365992002483381,1.0670896459263746,6.708964592637456

validate_tokenizer.py

@@ -0,0 +1,27 @@
+from pathlib import Path
+
+from tokenizers import Tokenizer
+
+
+root = Path(__file__).resolve().parent
+tokenizer_path = root / "tokenizer.json"
+tokenizer = Tokenizer.from_file(str(tokenizer_path))
+
+if tokenizer.get_vocab_size() != 65_536:
+    raise RuntimeError(f"Unexpected vocabulary size: {tokenizer.get_vocab_size()}")
+
+samples = [
+    "المملكة العربية السعودية",
+    "السلام عليكم ورحمة الله وبركاته",
+    "Artificial intelligence والذكاء الاصطناعي",
+    "ياهلا ومرحبا فيكم",
+]
+
+for text in samples:
+    encoding = tokenizer.encode(text, add_special_tokens=False)
+    decoded = tokenizer.decode(encoding.ids)
+    if decoded != text:
+        raise RuntimeError(f"Round-trip failure: {text!r} -> {decoded!r}")
+
+print("Tokenizer validation passed")
+print(f"Vocabulary size: {tokenizer.get_vocab_size():,}")