Datasets:

Djangodevreng
/

dgx-spark-benchmarks

@@ -21,18 +21,18 @@ configs:
 # DGX Spark LLM Arena benchmarks
-Reproducible LLM inference benchmarks on an NVIDIA DGX Spark (GB10, 128 GB unified memory). The suite defines eleven tests: six closed-loop (llama-benchy) and five open-loop (vllm bench serve). The results published here cover nine of them (five closed-loop, four open-loop); concurrency stress and rate sweep are not published yet. Quality scores come from vendor model cards and are NOT normalised: each score carries the eval harness it came from (e.g. MMLU-Pro vs classic MMLU), because these are different scales. Compare across models only within the same harness.
 - **Hardware:** NVIDIA DGX Spark (GB10, 128 GB unified memory)
 - **Creator:** Django de Vreng (https://djangodevreng.nl)
 - **Visualised:** https://djangodevreng.nl/arena/
 - **Raw runs:** https://github.com/djangodevreng/dgx-spark-benchmarks
-- **Last updated:** 2026-06-26
 ## Configs
-- `models`: one row per model (19 rows). Metadata, quality scores and the leaderboard composite scores per use-case preset.
-- `results`: long format, one row per measured (model, benchmark) pair. The suite has 11 benchmarks — 6 closed-loop (llama-benchy) and 5 open-loop (vllm bench serve) — of which 9 carry per-model figures here.
 ### Quality columns

 # DGX Spark LLM Arena benchmarks
+Reproducible LLM inference benchmarks on an NVIDIA DGX Spark (GB10, 128 GB unified memory). The suite defines eleven tests: six closed-loop (llama-benchy) and five open-loop (vllm bench serve). Results cover all eleven: the ten throughput tests under `results`, and the rate sweep under `rateSweep`, which reports sustained capacity in requests per second under a p95 TTFT ceiling of 2, 5 and 10 seconds (null on a threshold means the model never met it). Quality scores come from vendor model cards and are NOT normalised: each score carries the eval harness it came from (e.g. MMLU-Pro vs classic MMLU), because these are different scales. Compare across models only within the same harness.
 - **Hardware:** NVIDIA DGX Spark (GB10, 128 GB unified memory)
 - **Creator:** Django de Vreng (https://djangodevreng.nl)
 - **Visualised:** https://djangodevreng.nl/arena/
 - **Raw runs:** https://github.com/djangodevreng/dgx-spark-benchmarks
+- **Last updated:** 2026-08-17
 ## Configs
+- `models`: one row per model (38 rows). Metadata, quality scores and the leaderboard composite scores per use-case preset.
+- `results`: long format, one row per measured (model, benchmark) pair. The suite has 11 benchmarks — 6 closed-loop (llama-benchy) and 5 open-loop (vllm bench serve) — of which 10 carry per-model figures here.
 ### Quality columns

models.csv CHANGED Viewed

@@ -1,20 +1,39 @@
 model_id,name,vendor,architecture,params_b,vram_gb,context_k,precision,knowledge,knowledge_bench,science,science_bench,coding,coding_bench,quality_avg,size_bucket,score_aggregate,score_chat,score_agent,score_batch,score_reasoning,detail_url,markdown_url
-qwen-3.5-0.8b-bf16,Qwen-3.5 0.8B,alibaba,Hybrid,0.8,2,256,BF16,29.7,MMLU-Pro,11.9,GPQA,,,20.8,small,46.9,54.5,46.9,84.8,54.5,https://djangodevreng.nl/arena/qwen-3-5-0-8b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-0-8b-bf16.md
-qwen-3.5-2b-bf16,Qwen-3.5 2B,alibaba,Hybrid,2,4,256,BF16,55.3,MMLU-Pro,30.4,SuperGPQA,,,42.8,small,47.5,47.1,44,47.5,45.4,https://djangodevreng.nl/arena/qwen-3-5-2b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-2b-bf16.md
-ministral-3-3b-instruct-bf16,Ministral-3 3B,mistral,Dense,3,8,256,BF16,70.7,MMLU,53.4,GPQA-Diamond,54.8,LiveCodeBench,59.6,small,59.6,58.1,61.5,42,51.9,https://djangodevreng.nl/arena/ministral-3-3b-instruct-bf16/,https://djangodevreng.nl/arena/ministral-3-3b-instruct-bf16.md
-ministral-3-8b-instruct-bf16,Ministral-3 8B,mistral,Dense,8,18,256,BF16,76.1,MMLU,66.8,GPQA-Diamond,61.6,LiveCodeBench,68.2,mid,60.4,55,58.1,31.2,52,https://djangodevreng.nl/arena/ministral-3-8b-instruct-bf16/,https://djangodevreng.nl/arena/ministral-3-8b-instruct-bf16.md
-gemma-4-bf16,Gemma-4 26B-A4B,google,MoE,26,52,256,BF16,82.6,MMLU-Pro,82.3,GPQA-Diamond,77.1,LiveCodeBench v6,80.7,mid,68.9,60.4,68.7,28.5,56.3,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16-v23/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16-v23.md
-gemma-4-nvfp4,Gemma-4 26B-A4B,google,MoE,26,24,256,NVFP4,84.8,MMLU-Pro,79.9,GPQA-Diamond,79.8,LiveCodeBench v6,81.5,mid,71.8,63.8,70.6,33.9,56.9,https://djangodevreng.nl/arena/gemma-4-26b-a4b-nvfp4-v23/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-nvfp4-v23.md
-gemma-4-mtp,Gemma-4 26B-A4B,google,MoE,26,52,256,BF16 + MTP,82.6,MMLU-Pro,82.3,GPQA-Diamond,77.1,LiveCodeBench v6,80.7,mid,70.1,62.6,70.1,30.9,56.3,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-mtp-v23/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-mtp-v23.md
-nemotron-3-bf16,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,62,256,BF16,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,60.5,52.9,60,25.1,49.5,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-bf16/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-bf16.md
-nemotron-3-fp8,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,33,256,FP8,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,62.5,55.8,61,31.4,49.5,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-fp8/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-fp8.md
-nemotron-3-nvfp4,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,21,256,NVFP4,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,64.9,58.8,63.8,37.5,49.5,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-nvfp4/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-nvfp4.md
-qwen-3.6-fp8,Qwen-3.6 27B,alibaba,Hybrid,27,31,256,FP8,86.2,MMLU-Pro,87.8,GPQA-Diamond,83.9,LiveCodeBench v6,86,mid,71.9,63.2,71.6,25.6,60,https://djangodevreng.nl/arena/qwen-3-6-27b-fp8/,https://djangodevreng.nl/arena/qwen-3-6-27b-fp8.md
-qwen-3.6-35b-a3b-fp8,Qwen-3.6 35B-A3B,alibaba,MoE,35,38,256,FP8,85.2,MMLU-Pro,86,GPQA-Diamond,80.4,LiveCodeBench v6,83.9,large,73.1,65.3,72.8,33.2,58.5,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-fp8/,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-fp8.md
-qwen-3.6-35b-a3b-bf16,Qwen-3.6 35B-A3B,alibaba,MoE,35,70,256,BF16,85.2,MMLU-Pro,86,GPQA-Diamond,80.4,LiveCodeBench v6,83.9,large,70.7,62.3,70.9,27.5,60.5,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-bf16/,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-bf16.md
-nemotron-3-nano-4b-bf16,Nemotron-3-Nano 4B,nvidia,Dense,4,8,256,BF16,18.1,MMLU-Pro,51.3,GPQA-Diamond,51.8,LiveCodeBench,40.4,small,39.6,39.1,36.6,30.7,33.9,https://djangodevreng.nl/arena/nemotron-3-nano-4b-bf16/,https://djangodevreng.nl/arena/nemotron-3-nano-4b-bf16.md
-nemotron-3-super-nvfp4,Nemotron-3-Super 120B-A12B,nvidia,MoE,120,60,256,NVFP4,83.7,MMLU-Pro,79.2,GPQA,81.2,LiveCodeBench v5,81.4,large,68.2,60.2,68.4,25.2,58.7,https://djangodevreng.nl/arena/nemotron-3-super-120b-a12b-nvfp4/,https://djangodevreng.nl/arena/nemotron-3-super-120b-a12b-nvfp4.md
-gemma-4-31b-bf16,Gemma-4 31B,google,Dense,31,62,256,BF16,85.2,MMLU-Pro,84.3,GPQA-Diamond,80,LiveCodeBench v6,83.2,mid,68.4,59,68.4,21.7,59,https://djangodevreng.nl/arena/gemma-4-31b-it-bf16/,https://djangodevreng.nl/arena/gemma-4-31b-it-bf16.md
-mistral-small-3.2-24b-nvfp4,Mistral-Small 3.2 24B,mistral,Dense,24,16,128,NVFP4,80.5,MMLU,46.13,GPQA-Diamond,92.9,HumanEval+ pass@5,73.2,mid,61.9,54,62.5,19.2,51.1,https://djangodevreng.nl/arena/mistral-small-3-2-24b-it-nvfp4/,https://djangodevreng.nl/arena/mistral-small-3-2-24b-it-nvfp4.md
-qwen-3.5-9b-bf16,Qwen-3.5 9B,alibaba,Dense,9,18,256,BF16,82.5,MMLU-Pro,81.7,GPQA-Diamond,65.6,LiveCodeBench v6,76.6,mid,66.2,59.3,66.9,28.8,56.4,https://djangodevreng.nl/arena/qwen-3-5-9b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-9b-bf16.md
-gpt-oss-20b-mxfp4,gpt-oss-20b,openai,MoE,21,14,128,MXFP4,85.3,MMLU,71.5,GPQA-Diamond,,,78.4,mid,69.8,62.8,68.7,39,60.7,https://djangodevreng.nl/arena/gpt-oss-20b-mxfp4/,https://djangodevreng.nl/arena/gpt-oss-20b-mxfp4.md

 model_id,name,vendor,architecture,params_b,vram_gb,context_k,precision,knowledge,knowledge_bench,science,science_bench,coding,coding_bench,quality_avg,size_bucket,score_aggregate,score_chat,score_agent,score_batch,score_reasoning,detail_url,markdown_url
+gemma-4-26b-a4b-it-bf16,Gemma-4 26B-A4B,google,MoE,26,52,256,BF16,82.6,MMLU-Pro,82.3,GPQA-Diamond,77.1,LiveCodeBench v6,80.7,mid,65.7,58.5,66.3,27,56.3,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16.md
+gemma-4-26b-a4b-it-nvfp4,Gemma-4 26B-A4B,google,MoE,26,18,256,NVFP4,82.6,MMLU-Pro,82.3,GPQA-Diamond,77.1,LiveCodeBench v6,80.7,mid,68.1,61.4,70.3,32.8,58.4,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-nvfp4/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-nvfp4.md
+nemotron-3-bf16,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,62,256,BF16,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,57.8,51.2,58,24,50.3,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-bf16/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-bf16.md
+nemotron-3-fp8,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,33,256,FP8,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,59.6,54.4,60.2,29.2,51.7,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-fp8/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-fp8.md
+nemotron-3-nvfp4,Nemotron-3-Nano 30B-A3B,nvidia,MoE,30,21,256,NVFP4,77.3,MMLU-Pro,72.2,GPQA-Diamond,63.2,LiveCodeBench v5,70.9,mid,61.6,58.7,63.6,34.1,53.8,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-nvfp4/,https://djangodevreng.nl/arena/nemotron-3-nano-30b-a3b-nvfp4.md
+qwen-3.6-fp8,Qwen-3.6 27B,alibaba,Hybrid,27,31,256,FP8,86.2,MMLU-Pro,87.8,GPQA-Diamond,83.9,LiveCodeBench v6,86,mid,68.5,61,69.3,22.5,59.4,https://djangodevreng.nl/arena/qwen-3-6-27b-fp8/,https://djangodevreng.nl/arena/qwen-3-6-27b-fp8.md
+qwen-3.6-35b-a3b-fp8,Qwen-3.6 35B-A3B,alibaba,MoE,35,38,256,FP8,85.2,MMLU-Pro,86,GPQA-Diamond,80.4,LiveCodeBench v6,83.9,large,69.7,64.2,71.2,30.9,59.4,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-fp8/,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-fp8.md
+qwen-3.5-0.8b-bf16,Qwen-3.5 0.8B,alibaba,Hybrid,0.8,2,256,BF16,29.7,MMLU-Pro,11.9,GPQA,,,20.8,small,46.2,53.9,46.2,84.6,53.9,https://djangodevreng.nl/arena/qwen-3-5-0-8b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-0-8b-bf16.md
+nemotron-3-nano-4b-bf16,Nemotron-3-Nano 4B,nvidia,Dense,4,8,256,BF16,18.1,MMLU-Pro,51.3,GPQA-Diamond,51.8,LiveCodeBench,40.4,small,38.5,39.9,38.8,29,32.2,https://djangodevreng.nl/arena/nemotron-3-nano-4b-bf16/,https://djangodevreng.nl/arena/nemotron-3-nano-4b-bf16.md
+nemotron-3-super-nvfp4,Nemotron-3-Super 120B-A12B,nvidia,MoE,120,60,256,NVFP4,83.7,MMLU-Pro,79.2,GPQA,81.2,LiveCodeBench v5,81.4,large,65.3,58.3,66,23.4,56.3,https://djangodevreng.nl/arena/nemotron-3-super-120b-a12b-nvfp4/,https://djangodevreng.nl/arena/nemotron-3-super-120b-a12b-nvfp4.md
+qwen-3.6-35b-a3b-bf16,Qwen-3.6 35B-A3B,alibaba,MoE,35,70,256,BF16,85.2,MMLU-Pro,86,GPQA-Diamond,80.4,LiveCodeBench v6,83.9,large,68,60.6,68.7,26.9,58,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-bf16/,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-bf16.md
+gemma-4-31b-bf16,Gemma-4 31B,google,Dense,31,62,256,BF16,85.2,MMLU-Pro,84.3,GPQA-Diamond,80,LiveCodeBench v6,83.2,mid,65.5,56.6,65.7,20.7,56.6,https://djangodevreng.nl/arena/gemma-4-31b-it-bf16/,https://djangodevreng.nl/arena/gemma-4-31b-it-bf16.md
+qwen-3.5-2b-bf16,Qwen-3.5 2B,alibaba,Hybrid,2,4,256,BF16,55.3,MMLU-Pro,30.4,SuperGPQA,,,42.8,small,46.7,48.5,46,46.2,44.2,https://djangodevreng.nl/arena/qwen-3-5-2b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-2b-bf16.md
+ministral-3-3b-instruct-bf16,Ministral-3 3B,mistral,Dense,3,8,256,BF16,70.7,MMLU,53.4,GPQA-Diamond,54.8,LiveCodeBench,59.6,small,55.6,60.4,57.6,32.9,45.5,https://djangodevreng.nl/arena/ministral-3-3b-instruct-bf16/,https://djangodevreng.nl/arena/ministral-3-3b-instruct-bf16.md
+ministral-3-8b-instruct-bf16,Ministral-3 8B,mistral,Dense,8,18,256,BF16,76.1,MMLU,66.8,GPQA-Diamond,61.6,LiveCodeBench,68.2,mid,58.6,55.7,58.3,30.3,50,https://djangodevreng.nl/arena/ministral-3-8b-instruct-bf16/,https://djangodevreng.nl/arena/ministral-3-8b-instruct-bf16.md
+gemma-4-bf16,Gemma-4 26B-A4B,google,MoE,26,52,256,BF16,82.6,MMLU-Pro,82.3,GPQA-Diamond,77.1,LiveCodeBench v6,80.7,mid,65.7,58.8,66.3,27.1,56.3,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16-v23/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-bf16-v23.md
+gemma-4-nvfp4,Gemma-4 26B-A4B,google,MoE,26,24,256,NVFP4,84.8,MMLU-Pro,79.9,GPQA-Diamond,79.8,LiveCodeBench v6,81.5,mid,68.7,62.2,70.5,33,58.9,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-nvfp4-v23/,https://djangodevreng.nl/arena/gemma-4-26b-a4b-it-nvfp4-v23.md
+gpt-oss-20b-mxfp4,gpt-oss-20b,openai,MoE,21,14,128,MXFP4,85.3,MMLU,71.5,GPQA-Diamond,,,78.4,mid,67.2,62.4,70.1,37.2,58.9,https://djangodevreng.nl/arena/gpt-oss-20b-mxfp4/,https://djangodevreng.nl/arena/gpt-oss-20b-mxfp4.md
+qwen-3.5-9b-bf16,Qwen-3.5 9B,alibaba,Dense,9,18,256,BF16,82.5,MMLU-Pro,81.7,GPQA-Diamond,65.6,LiveCodeBench v6,76.6,mid,63.4,58,65.1,27.3,54.1,https://djangodevreng.nl/arena/qwen-3-5-9b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-9b-bf16.md
+gemma-4-31b-it-nvfp4,Gemma-4 31B,google,Dense,31,22,256,NVFP4,85.2,MMLU-Pro,84.3,GPQA-Diamond,80,LiveCodeBench v6,83.2,mid,66,57.6,66.4,21.7,57.1,https://djangodevreng.nl/arena/gemma-4-31b-it-nvfp4/,https://djangodevreng.nl/arena/gemma-4-31b-it-nvfp4.md
+gemma-4-e2b-it-bf16,Gemma-4 E2B,google,Dense,2.3,5,128,BF16,60,MMLU-Pro,43.4,GPQA-Diamond,44,LiveCodeBench v6,49.1,small,49.7,48.1,51.6,44,46.3,https://djangodevreng.nl/arena/gemma-4-e2b-it-bf16/,https://djangodevreng.nl/arena/gemma-4-e2b-it-bf16.md
+gemma-4-e4b-it-bf16,Gemma-4 E4B,google,Dense,4.5,9,128,BF16,69.4,MMLU-Pro,58.6,GPQA-Diamond,52,LiveCodeBench v6,60,small,51.9,47.7,52.6,27.9,45.2,https://djangodevreng.nl/arena/gemma-4-e4b-it-bf16/,https://djangodevreng.nl/arena/gemma-4-e4b-it-bf16.md
+granite-4-1-8b-bf16,Granite-4.1 8B,ibm,Dense,8,16,128,BF16,55.99,MMLU-Pro,41.96,GPQA,85.37,HumanEval,61.1,mid,50.4,44.1,51,22,43.3,https://djangodevreng.nl/arena/granite-4-1-8b-bf16/,https://djangodevreng.nl/arena/granite-4-1-8b-bf16.md
+kat-coder-v2-5-bf16,KAT-Coder V2.5,kwaipilot,MoE,35,70,256,BF16,,,,,,,0,large,2.5,4.1,2.7,7.9,1.9,https://djangodevreng.nl/arena/kat-coder-v2-5-bf16/,https://djangodevreng.nl/arena/kat-coder-v2-5-bf16.md
+lfm2-5-2-6b-bf16,LFM2.5 2.6B,liquidai,Hybrid,2.7,5,32,BF16,,,,,59.41,LiveCodeBench v6,59.4,small,58.2,62,58.4,45.5,51.6,https://djangodevreng.nl/arena/lfm2-5-2-6b-bf16/,https://djangodevreng.nl/arena/lfm2-5-2-6b-bf16.md
+mistral-small-4-119b-nvfp4,Mistral-Small 4 119B,mistral,MoE,119,83,256,NVFP4,,,71.2,GPQA-Diamond,,,71.2,large,58,51.9,59,23.2,49,https://djangodevreng.nl/arena/mistral-small-4-119b-nvfp4/,https://djangodevreng.nl/arena/mistral-small-4-119b-nvfp4.md
+muse-glimmer-30b-bf16,Muse-Glimmer 30B,meta,Dense,30,60,128,BF16,,,83.5,GPQA-Diamond,,,83.5,mid,66.3,58,66.5,22.4,56.8,https://djangodevreng.nl/arena/muse-glimmer-30b-bf16/,https://djangodevreng.nl/arena/muse-glimmer-30b-bf16.md
+muse-glimmer-30b-bf16-spec,Muse-Glimmer 30B,meta,Dense,30,60,128,BF16,,,83.5,GPQA-Diamond,,,83.5,mid,67.1,60.4,68.9,23.9,57.2,https://djangodevreng.nl/arena/muse-glimmer-30b-bf16-spec/,https://djangodevreng.nl/arena/muse-glimmer-30b-bf16-spec.md
+nemotron-3-nano-4b-fp8,Nemotron-3-Nano 4B,nvidia,Dense,4,4,256,FP8,18.1,MMLU-Pro,51.3,GPQA-Diamond,51.8,LiveCodeBench,40.4,small,42,46.1,41.7,37.8,35.3,https://djangodevreng.nl/arena/nemotron-3-nano-4b-fp8/,https://djangodevreng.nl/arena/nemotron-3-nano-4b-fp8.md
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,Nemotron-3 Nano Omni 30B-A3B,nvidia,MoE,30,60,256,BF16,,,,,,,0,mid,2.5,3.9,2.6,8.1,2.4,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-bf16/,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-bf16.md
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,Nemotron-3 Nano Omni 30B-A3B,nvidia,MoE,30,33,256,FP8,,,,,,,0,mid,4.6,7.9,5.2,14.1,3.9,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-fp8/,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-fp8.md
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,Nemotron-3 Nano Omni 30B-A3B,nvidia,MoE,30,21,256,NVFP4,,,,,,,0,mid,6.4,12,6.9,18.6,7.9,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4/,https://djangodevreng.nl/arena/nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4.md
+nemotron-cascade-2-30b-a3b-bf16,Nemotron-Cascade-2 30B-A3B,nvidia,MoE,30,60,1000,BF16,79.8,MMLU-Pro,76.1,GPQA-Diamond,87.2,LiveCodeBench v6,81,mid,65.7,58.2,66,26.1,56,https://djangodevreng.nl/arena/nemotron-cascade-2-30b-a3b-bf16/,https://djangodevreng.nl/arena/nemotron-cascade-2-30b-a3b-bf16.md
+qwen-3-5-4b-bf16,Qwen-3.5 4B,alibaba,Hybrid,4,8,256,BF16,79.1,MMLU-Pro,76.2,GPQA-Diamond,55.8,LiveCodeBench v6,70.4,small,61.7,59.5,65.1,33.3,51.7,https://djangodevreng.nl/arena/qwen-3-5-4b-bf16/,https://djangodevreng.nl/arena/qwen-3-5-4b-bf16.md
+qwen-3-6-27b-bf16,Qwen-3.6 27B,alibaba,Hybrid,27,54,256,BF16,86.2,MMLU-Pro,87.8,GPQA-Diamond,83.9,LiveCodeBench v6,86,mid,68.1,59.6,68.3,22.4,58.4,https://djangodevreng.nl/arena/qwen-3-6-27b-bf16/,https://djangodevreng.nl/arena/qwen-3-6-27b-bf16.md
+qwen-3-6-35b-a3b-nvfp4,Qwen-3.6 35B-A3B,alibaba,MoE,35,24,256,NVFP4,85.2,MMLU-Pro,86,GPQA-Diamond,80.4,LiveCodeBench v6,83.9,large,71.1,66.8,73,35,60,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-nvfp4/,https://djangodevreng.nl/arena/qwen-3-6-35b-a3b-nvfp4.md
+qwen-3-8-27b-bf16,Qwen-3.8 27B,alibaba,Dense,27,54,256,BF16,,,89.2,GPQA-Diamond,90.3,LiveCodeBench v6,89.8,mid,71.1,62.1,71.6,23.2,61,https://djangodevreng.nl/arena/qwen-3-8-27b-bf16/,https://djangodevreng.nl/arena/qwen-3-8-27b-bf16.md
+qwen-3-8-27b-fp8,Qwen-3.8 27B,alibaba,Dense,27,30,256,FP8,,,89.2,GPQA-Diamond,90.3,LiveCodeBench v6,89.8,mid,71.5,63.6,72.8,23.4,61.9,https://djangodevreng.nl/arena/qwen-3-8-27b-fp8/,https://djangodevreng.nl/arena/qwen-3-8-27b-fp8.md

results.csv CHANGED Viewed

@@ -1,163 +1,381 @@
 model_id,benchmark_id,benchmark_num,benchmark_title,mode,tokens_per_sec_per_user,tokens_per_sec_total,ttft_ms
-qwen-3.5-0.8b-bf16,chat,01,01-chat,closed,102,669,260
-qwen-3.5-0.8b-bf16,rag-8k,02,02-rag-8k,closed,73,565,1720
-qwen-3.5-0.8b-bf16,long-output,03,03-long-output,closed,106,588,120
-qwen-3.5-0.8b-bf16,multi-turn,04,04-multi-turn,closed,91,729,470
-qwen-3.5-0.8b-bf16,big-context,05,05-big-context,closed,36,182,6180
-qwen-3.5-0.8b-bf16,office-baseline,07,07-office-baseline,open,227,227,224
-qwen-3.5-0.8b-bf16,sharegpt,08,08-sharegpt,open,22,22,43
-qwen-3.5-0.8b-bf16,reasoning,09,09-reasoning,open,63,63,96
-qwen-3.5-0.8b-bf16,monday-peak,10,10-monday-peak,open,281,281,269
-qwen-3.5-2b-bf16,chat,01,01-chat,closed,50,329,390
-qwen-3.5-2b-bf16,rag-8k,02,02-rag-8k,closed,39,283,2740
-qwen-3.5-2b-bf16,long-output,03,03-long-output,closed,51,215,190
-qwen-3.5-2b-bf16,multi-turn,04,04-multi-turn,closed,48,379,740
-qwen-3.5-2b-bf16,big-context,05,05-big-context,closed,22,115,9430
-qwen-3.5-2b-bf16,office-baseline,07,07-office-baseline,open,136,136,366
-qwen-3.5-2b-bf16,sharegpt,08,08-sharegpt,open,18,18,76
-qwen-3.5-2b-bf16,reasoning,09,09-reasoning,open,28,28,169
-qwen-3.5-2b-bf16,monday-peak,10,10-monday-peak,open,180,180,470
-ministral-3-3b-instruct-bf16,chat,01,01-chat,closed,49,339,490
-ministral-3-3b-instruct-bf16,rag-8k,02,02-rag-8k,closed,26,219,3340
-ministral-3-3b-instruct-bf16,long-output,03,03-long-output,closed,53,340,190
-ministral-3-3b-instruct-bf16,multi-turn,04,04-multi-turn,closed,43,401,820
-ministral-3-3b-instruct-bf16,big-context,05,05-big-context,closed,10,61,15600
-ministral-3-3b-instruct-bf16,office-baseline,07,07-office-baseline,open,122,122,407
 ministral-3-3b-instruct-bf16,sharegpt,08,08-sharegpt,open,18,18,69
-ministral-3-3b-instruct-bf16,reasoning,09,09-reasoning,open,20,20,161
-ministral-3-3b-instruct-bf16,monday-peak,10,10-monday-peak,open,126,126,514
-ministral-3-8b-instruct-bf16,chat,01,01-chat,closed,23,160,960
-ministral-3-8b-instruct-bf16,rag-8k,02,02-rag-8k,closed,14,122,5920
-ministral-3-8b-instruct-bf16,long-output,03,03-long-output,closed,24,73,320
-ministral-3-8b-instruct-bf16,multi-turn,04,04-multi-turn,closed,21,195,1660
-ministral-3-8b-instruct-bf16,big-context,05,05-big-context,closed,6,38,25320
-ministral-3-8b-instruct-bf16,office-baseline,07,07-office-baseline,open,81,81,774
-ministral-3-8b-instruct-bf16,sharegpt,08,08-sharegpt,open,13,13,137
-ministral-3-8b-instruct-bf16,reasoning,09,09-reasoning,open,9,9,264
-ministral-3-8b-instruct-bf16,monday-peak,10,10-monday-peak,open,78,78,764
-gemma-4-bf16,chat,01,01-chat,closed,11,91,1343
-gemma-4-bf16,rag-8k,02,02-rag-8k,closed,10,78,8488
-gemma-4-bf16,long-output,03,03-long-output,closed,12,87,491
-gemma-4-bf16,multi-turn,04,04-multi-turn,closed,11,98,2155
-gemma-4-bf16,big-context,05,05-big-context,closed,5,28,39281
-gemma-4-bf16,office-baseline,07,07-office-baseline,open,34,34,1330
-gemma-4-bf16,sharegpt,08,08-sharegpt,open,8,8,327
-gemma-4-bf16,monday-peak,10,10-monday-peak,open,44,44,1178
-gemma-4-nvfp4,chat,01,01-chat,closed,22,151,1138
-gemma-4-nvfp4,rag-8k,02,02-rag-8k,closed,16,119,8309
-gemma-4-nvfp4,long-output,03,03-long-output,closed,24,121,369
-gemma-4-nvfp4,multi-turn,04,04-multi-turn,closed,20,183,1966
-gemma-4-nvfp4,big-context,05,05-big-context,closed,7,34,38575
-gemma-4-nvfp4,office-baseline,07,07-office-baseline,open,82,82,1051
-gemma-4-nvfp4,sharegpt,08,08-sharegpt,open,13,13,146
-gemma-4-nvfp4,monday-peak,10,10-monday-peak,open,73,73,967
-gemma-4-mtp,chat,01,01-chat,closed,18,139,1400
-gemma-4-mtp,rag-8k,02,02-rag-8k,closed,13,97,9519
-gemma-4-mtp,long-output,03,03-long-output,closed,18,128,564
-gemma-4-mtp,multi-turn,04,04-multi-turn,closed,17,143,2368
-gemma-4-mtp,big-context,05,05-big-context,closed,6,28,45640
-gemma-4-mtp,office-baseline,07,07-office-baseline,open,48,48,1608
-gemma-4-mtp,sharegpt,08,08-sharegpt,open,11,11,409
-gemma-4-mtp,monday-peak,10,10-monday-peak,open,53,53,1684
-nemotron-3-bf16,chat,01,01-chat,closed,8,76,1330
-nemotron-3-bf16,rag-8k,02,02-rag-8k,closed,8,69,6430
-nemotron-3-bf16,long-output,03,03-long-output,closed,10,67,550
-nemotron-3-bf16,multi-turn,04,04-multi-turn,closed,8,73,2000
-nemotron-3-bf16,big-context,05,05-big-context,closed,6,39,19820
-nemotron-3-bf16,office-baseline,07,07-office-baseline,open,29,29,1229
-nemotron-3-bf16,sharegpt,08,08-sharegpt,open,8,8,433
-nemotron-3-bf16,monday-peak,10,10-monday-peak,open,40,40,1130
-nemotron-3-fp8,chat,01,01-chat,closed,15,138,960
-nemotron-3-fp8,rag-8k,02,02-rag-8k,closed,14,119,5110
-nemotron-3-fp8,long-output,03,03-long-output,closed,18,95,450
-nemotron-3-fp8,multi-turn,04,04-multi-turn,closed,15,134,1550
-nemotron-3-fp8,big-context,05,05-big-context,closed,9,55,16890
-nemotron-3-fp8,office-baseline,07,07-office-baseline,open,72,72,732
-nemotron-3-fp8,sharegpt,08,08-sharegpt,open,11,11,220
-nemotron-3-fp8,monday-peak,10,10-monday-peak,open,70,70,757
-nemotron-3-nvfp4,chat,01,01-chat,closed,23,202,950
-nemotron-3-nvfp4,rag-8k,02,02-rag-8k,closed,20,167,4010
-nemotron-3-nvfp4,long-output,03,03-long-output,closed,25,172,360
-nemotron-3-nvfp4,multi-turn,04,04-multi-turn,closed,22,193,1360
-nemotron-3-nvfp4,big-context,05,05-big-context,closed,13,78,12710
-nemotron-3-nvfp4,office-baseline,07,07-office-baseline,open,89,89,618
-nemotron-3-nvfp4,sharegpt,08,08-sharegpt,open,13,13,157
-nemotron-3-nvfp4,monday-peak,10,10-monday-peak,open,94,94,687
-qwen-3.6-fp8,chat,01,01-chat,closed,7,69,4480
-qwen-3.6-fp8,rag-8k,02,02-rag-8k,closed,6,45,23310
-qwen-3.6-fp8,long-output,03,03-long-output,closed,7,46,2090
-qwen-3.6-fp8,multi-turn,04,04-multi-turn,closed,7,63,6250
-qwen-3.6-fp8,big-context,05,05-big-context,closed,3,15,76490
-qwen-3.6-fp8,office-baseline,07,07-office-baseline,open,6,6,14093
-qwen-3.6-fp8,sharegpt,08,08-sharegpt,open,7,7,511
-qwen-3.6-fp8,monday-peak,10,10-monday-peak,open,25,25,3882
-qwen-3.6-35b-a3b-fp8,chat,01,01-chat,closed,15,148,1210
-qwen-3.6-35b-a3b-fp8,rag-8k,02,02-rag-8k,closed,13,110,8050
-qwen-3.6-35b-a3b-fp8,long-output,03,03-long-output,closed,17,128,530
-qwen-3.6-35b-a3b-fp8,multi-turn,04,04-multi-turn,closed,15,139,2130
-qwen-3.6-35b-a3b-fp8,big-context,05,05-big-context,closed,8,43,26880
-qwen-3.6-35b-a3b-fp8,office-baseline,07,07-office-baseline,open,54,54,1229
-qwen-3.6-35b-a3b-fp8,sharegpt,08,08-sharegpt,open,11,11,205
-qwen-3.6-35b-a3b-fp8,monday-peak,10,10-monday-peak,open,59,59,1352
-qwen-3.6-35b-a3b-bf16,chat,01,01-chat,closed,8,83,1780
-qwen-3.6-35b-a3b-bf16,rag-8k,02,02-rag-8k,closed,7,64,11950
-qwen-3.6-35b-a3b-bf16,long-output,03,03-long-output,closed,10,76,730
-qwen-3.6-35b-a3b-bf16,multi-turn,04,04-multi-turn,closed,8,80,3220
-qwen-3.6-35b-a3b-bf16,big-context,05,05-big-context,closed,5,27,39370
-qwen-3.6-35b-a3b-bf16,office-baseline,07,07-office-baseline,open,21,21,3100
-qwen-3.6-35b-a3b-bf16,sharegpt,08,08-sharegpt,open,8,8,390
-qwen-3.6-35b-a3b-bf16,reasoning,09,09-reasoning,open,4,4,741
-qwen-3.6-35b-a3b-bf16,monday-peak,10,10-monday-peak,open,34,34,2153
-nemotron-3-nano-4b-bf16,chat,01,01-chat,closed,25,224,750
-nemotron-3-nano-4b-bf16,rag-8k,02,02-rag-8k,closed,21,166,5090
-nemotron-3-nano-4b-bf16,long-output,03,03-long-output,closed,26,96,300
-nemotron-3-nano-4b-bf16,multi-turn,04,04-multi-turn,closed,24,221,1370
-nemotron-3-nano-4b-bf16,big-context,05,05-big-context,closed,13,67,16530
-nemotron-3-nano-4b-bf16,office-baseline,07,07-office-baseline,open,88,88,706
-nemotron-3-nano-4b-bf16,sharegpt,08,08-sharegpt,open,13,13,134
-nemotron-3-nano-4b-bf16,reasoning,09,09-reasoning,open,11,11,257
-nemotron-3-nano-4b-bf16,monday-peak,10,10-monday-peak,open,100,100,775
-nemotron-3-super-nvfp4,chat,01,01-chat,closed,8,75,4600
-nemotron-3-super-nvfp4,rag-8k,02,02-rag-8k,closed,7,50,26510
-nemotron-3-super-nvfp4,long-output,03,03-long-output,closed,9,61,2050
-nemotron-3-super-nvfp4,multi-turn,04,04-multi-turn,closed,8,69,7780
-nemotron-3-super-nvfp4,big-context,05,05-big-context,closed,4,17,79790
-nemotron-3-super-nvfp4,office-baseline,07,07-office-baseline,open,7,7,247722
-nemotron-3-super-nvfp4,sharegpt,08,08-sharegpt,open,7,7,1082
-nemotron-3-super-nvfp4,reasoning,09,09-reasoning,open,4,4,281808
-nemotron-3-super-nvfp4,monday-peak,10,10-monday-peak,open,26,26,33368
-gemma-4-31b-bf16,chat,01,01-chat,closed,3,20,7190
-gemma-4-31b-bf16,rag-8k,02,02-rag-8k,closed,3,19,44190
-gemma-4-31b-bf16,long-output,03,03-long-output,closed,4,19,2090
-gemma-4-31b-bf16,multi-turn,04,04-multi-turn,closed,3,22,11760
-gemma-4-31b-bf16,big-context,05,05-big-context,closed,1,7,177230
-gemma-4-31b-bf16,office-baseline,07,07-office-baseline,open,4,4,164379
-gemma-4-31b-bf16,sharegpt,08,08-sharegpt,open,3,3,903
-gemma-4-31b-bf16,reasoning,09,09-reasoning,open,2,2,1846
-gemma-4-31b-bf16,monday-peak,10,10-monday-peak,open,15,15,4970
-mistral-small-3.2-24b-nvfp4,chat,01,01-chat,closed,13,63,7010
-mistral-small-3.2-24b-nvfp4,rag-8k,02,02-rag-8k,closed,5,17,39890
-mistral-small-3.2-24b-nvfp4,long-output,03,03-long-output,closed,15,80,5200
-mistral-small-3.2-24b-nvfp4,multi-turn,04,04-multi-turn,closed,13,94,12380
-mistral-small-3.2-24b-nvfp4,big-context,05,05-big-context,closed,2,5,127730
-mistral-small-3.2-24b-nvfp4,office-baseline,07,07-office-baseline,open,5,5,86740
-mistral-small-3.2-24b-nvfp4,sharegpt,08,08-sharegpt,open,22,22,1004
-mistral-small-3.2-24b-nvfp4,monday-peak,10,10-monday-peak,open,30,30,4842
-qwen-3.5-9b-bf16,chat,01,01-chat,closed,13,126,1470
-qwen-3.5-9b-bf16,rag-8k,02,02-rag-8k,closed,10,87,10660
-qwen-3.5-9b-bf16,long-output,03,03-long-output,closed,13,127,500
-qwen-3.5-9b-bf16,multi-turn,04,04-multi-turn,closed,12,116,2790
-qwen-3.5-9b-bf16,big-context,05,05-big-context,closed,6,32,34640
-qwen-3.5-9b-bf16,office-baseline,07,07-office-baseline,open,48,48,1701
-qwen-3.5-9b-bf16,sharegpt,08,08-sharegpt,open,9,9,244
-qwen-3.5-9b-bf16,reasoning,09,09-reasoning,open,6,6,518
-qwen-3.5-9b-bf16,monday-peak,10,10-monday-peak,open,54,54,1617
-gpt-oss-20b-mxfp4,chat,01,01-chat,closed,26,173,910
-gpt-oss-20b-mxfp4,rag-8k,02,02-rag-8k,closed,22,164,6200
-gpt-oss-20b-mxfp4,long-output,03,03-long-output,closed,25,135,320
-gpt-oss-20b-mxfp4,multi-turn,04,04-multi-turn,closed,25,151,1610
-gpt-oss-20b-mxfp4,big-context,05,05-big-context,closed,10,45,26660
-gpt-oss-20b-mxfp4,office-baseline,07,07-office-baseline,open,128,128,735
-gpt-oss-20b-mxfp4,sharegpt,08,08-sharegpt,open,17,17,118
-gpt-oss-20b-mxfp4,reasoning,09,09-reasoning,open,12,12,244
-gpt-oss-20b-mxfp4,monday-peak,10,10-monday-peak,open,100,100,745

 model_id,benchmark_id,benchmark_num,benchmark_title,mode,tokens_per_sec_per_user,tokens_per_sec_total,ttft_ms
+gemma-4-26b-a4b-it-bf16,chat,01,01-chat,closed,11,85,1473
+gemma-4-26b-a4b-it-bf16,rag-8k,02,02-rag-8k,closed,9,75,8806
+gemma-4-26b-a4b-it-bf16,long-output,03,03-long-output,closed,12,81,524
+gemma-4-26b-a4b-it-bf16,multi-turn,04,04-multi-turn,closed,10,90,2405
+gemma-4-26b-a4b-it-bf16,big-context,05,05-big-context,closed,9,80,4327
+gemma-4-26b-a4b-it-bf16,concurrency-stress,06,06-concurrency-stress,closed,3,29,77388
+gemma-4-26b-a4b-it-bf16,office-baseline,07,07-office-baseline,open,31,31,1435
+gemma-4-26b-a4b-it-bf16,sharegpt,08,08-sharegpt,open,8,8,364
+gemma-4-26b-a4b-it-bf16,reasoning,09,09-reasoning,open,5,5,602
+gemma-4-26b-a4b-it-bf16,monday-peak,10,10-monday-peak,open,41,41,1246
+gemma-4-26b-a4b-it-nvfp4,chat,01,01-chat,closed,21,135,1347
+gemma-4-26b-a4b-it-nvfp4,rag-8k,02,02-rag-8k,closed,16,122,8118
+gemma-4-26b-a4b-it-nvfp4,long-output,03,03-long-output,closed,23,169,353
+gemma-4-26b-a4b-it-nvfp4,multi-turn,04,04-multi-turn,closed,20,175,2077
+gemma-4-26b-a4b-it-nvfp4,big-context,05,05-big-context,closed,17,139,3894
+gemma-4-26b-a4b-it-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,4,33,74374
+gemma-4-26b-a4b-it-nvfp4,office-baseline,07,07-office-baseline,open,82,82,1096
+gemma-4-26b-a4b-it-nvfp4,sharegpt,08,08-sharegpt,open,13,13,148
+gemma-4-26b-a4b-it-nvfp4,reasoning,09,09-reasoning,open,9,9,343
+gemma-4-26b-a4b-it-nvfp4,monday-peak,10,10-monday-peak,open,69,69,969
+nemotron-3-bf16,chat,01,01-chat,closed,8,72,1506
+nemotron-3-bf16,rag-8k,02,02-rag-8k,closed,7,68,6283
+nemotron-3-bf16,long-output,03,03-long-output,closed,8,66,612
+nemotron-3-bf16,multi-turn,04,04-multi-turn,closed,8,74,2119
+nemotron-3-bf16,big-context,05,05-big-context,closed,8,69,3575
+nemotron-3-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,48,36554
+nemotron-3-bf16,office-baseline,07,07-office-baseline,open,28,28,1232
+nemotron-3-bf16,sharegpt,08,08-sharegpt,open,8,8,435
+nemotron-3-bf16,reasoning,09,09-reasoning,open,6,6,662
+nemotron-3-bf16,monday-peak,10,10-monday-peak,open,39,39,1113
+nemotron-3-fp8,chat,01,01-chat,closed,14,130,1959
+nemotron-3-fp8,rag-8k,02,02-rag-8k,closed,13,111,5224
+nemotron-3-fp8,long-output,03,03-long-output,closed,16,117,733
+nemotron-3-fp8,multi-turn,04,04-multi-turn,closed,14,129,2159
+nemotron-3-fp8,big-context,05,05-big-context,closed,13,116,3099
+nemotron-3-fp8,concurrency-stress,06,06-concurrency-stress,closed,6,68,29327
+nemotron-3-fp8,office-baseline,07,07-office-baseline,open,54,54,1022
+nemotron-3-fp8,sharegpt,08,08-sharegpt,open,11,11,217
+nemotron-3-fp8,reasoning,09,09-reasoning,open,9,9,620
+nemotron-3-fp8,monday-peak,10,10-monday-peak,open,60,60,986
+nemotron-3-nvfp4,chat,01,01-chat,closed,22,206,1195
+nemotron-3-nvfp4,rag-8k,02,02-rag-8k,closed,19,151,6580
+nemotron-3-nvfp4,long-output,03,03-long-output,closed,25,192,445
+nemotron-3-nvfp4,multi-turn,04,04-multi-turn,closed,21,196,2066
+nemotron-3-nvfp4,big-context,05,05-big-context,closed,19,157,3553
+nemotron-3-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,7,61,38514
+nemotron-3-nvfp4,office-baseline,07,07-office-baseline,open,89,89,884
+nemotron-3-nvfp4,sharegpt,08,08-sharegpt,open,15,15,138
+nemotron-3-nvfp4,reasoning,09,09-reasoning,open,13,13,455
+nemotron-3-nvfp4,monday-peak,10,10-monday-peak,open,84,84,903
+qwen-3.6-fp8,chat,01,01-chat,closed,7,65,11411
+qwen-3.6-fp8,rag-8k,02,02-rag-8k,closed,4,28,70196
+qwen-3.6-fp8,long-output,03,03-long-output,closed,7,53,3715
+qwen-3.6-fp8,multi-turn,04,04-multi-turn,closed,6,53,19440
+qwen-3.6-fp8,big-context,05,05-big-context,closed,5,29,36643
+qwen-3.6-fp8,concurrency-stress,06,06-concurrency-stress,closed,1,7,411608
+qwen-3.6-fp8,office-baseline,07,07-office-baseline,open,3,3,419516
+qwen-3.6-fp8,sharegpt,08,08-sharegpt,open,7,7,760
+qwen-3.6-fp8,reasoning,09,09-reasoning,open,4,4,2947
+qwen-3.6-fp8,monday-peak,10,10-monday-peak,open,16,16,8869
+qwen-3.6-35b-a3b-fp8,chat,01,01-chat,closed,15,150,1652
+qwen-3.6-35b-a3b-fp8,rag-8k,02,02-rag-8k,closed,13,101,10739
+qwen-3.6-35b-a3b-fp8,long-output,03,03-long-output,closed,18,136,654
+qwen-3.6-35b-a3b-fp8,multi-turn,04,04-multi-turn,closed,15,141,2995
+qwen-3.6-35b-a3b-fp8,big-context,05,05-big-context,closed,13,106,5371
+qwen-3.6-35b-a3b-fp8,concurrency-stress,06,06-concurrency-stress,closed,5,37,66864
+qwen-3.6-35b-a3b-fp8,office-baseline,07,07-office-baseline,open,54,54,1656
+qwen-3.6-35b-a3b-fp8,sharegpt,08,08-sharegpt,open,13,13,188
+qwen-3.6-35b-a3b-fp8,reasoning,09,09-reasoning,open,7,7,493
+qwen-3.6-35b-a3b-fp8,monday-peak,10,10-monday-peak,open,55,55,1542
+qwen-3.5-0.8b-bf16,chat,01,01-chat,closed,102,564,234
+qwen-3.5-0.8b-bf16,rag-8k,02,02-rag-8k,closed,74,585,1630
+qwen-3.5-0.8b-bf16,long-output,03,03-long-output,closed,106,484,119
+qwen-3.5-0.8b-bf16,multi-turn,04,04-multi-turn,closed,96,757,417
+qwen-3.5-0.8b-bf16,big-context,05,05-big-context,closed,80,636,809
+qwen-3.5-0.8b-bf16,concurrency-stress,06,06-concurrency-stress,closed,23,196,11497
+qwen-3.5-0.8b-bf16,office-baseline,07,07-office-baseline,open,227,227,213
+qwen-3.5-0.8b-bf16,sharegpt,08,08-sharegpt,open,22,22,36
+qwen-3.5-0.8b-bf16,reasoning,09,09-reasoning,open,63,63,82
+qwen-3.5-0.8b-bf16,monday-peak,10,10-monday-peak,open,294,294,252
+nemotron-3-nano-4b-bf16,chat,01,01-chat,closed,24,224,883
+nemotron-3-nano-4b-bf16,rag-8k,02,02-rag-8k,closed,20,161,5546
+nemotron-3-nano-4b-bf16,long-output,03,03-long-output,closed,25,158,345
+nemotron-3-nano-4b-bf16,multi-turn,04,04-multi-turn,closed,23,211,1458
+nemotron-3-nano-4b-bf16,big-context,05,05-big-context,closed,20,169,2756
+nemotron-3-nano-4b-bf16,concurrency-stress,06,06-concurrency-stress,closed,7,66,34588
+nemotron-3-nano-4b-bf16,office-baseline,07,07-office-baseline,open,88,88,800
+nemotron-3-nano-4b-bf16,sharegpt,08,08-sharegpt,open,13,13,129
+nemotron-3-nano-4b-bf16,reasoning,09,09-reasoning,open,10,10,281
+nemotron-3-nano-4b-bf16,monday-peak,10,10-monday-peak,open,89,89,823
+nemotron-3-super-nvfp4,chat,01,01-chat,closed,8,73,6115
+nemotron-3-super-nvfp4,rag-8k,02,02-rag-8k,closed,6,44,34652
+nemotron-3-super-nvfp4,long-output,03,03-long-output,closed,9,60,2396
+nemotron-3-super-nvfp4,multi-turn,04,04-multi-turn,closed,8,64,10061
+nemotron-3-super-nvfp4,big-context,05,05-big-context,closed,7,46,18265
+nemotron-3-super-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,3,13,198114
+nemotron-3-super-nvfp4,office-baseline,07,07-office-baseline,open,5,5,339387
+nemotron-3-super-nvfp4,sharegpt,08,08-sharegpt,open,7,7,1089
+nemotron-3-super-nvfp4,reasoning,09,09-reasoning,open,4,4,284459
+nemotron-3-super-nvfp4,monday-peak,10,10-monday-peak,open,24,24,38453
+qwen-3.6-35b-a3b-bf16,chat,01,01-chat,closed,9,85,1903
+qwen-3.6-35b-a3b-bf16,rag-8k,02,02-rag-8k,closed,8,68,11775
+qwen-3.6-35b-a3b-bf16,long-output,03,03-long-output,closed,10,79,844
+qwen-3.6-35b-a3b-bf16,multi-turn,04,04-multi-turn,closed,9,82,3178
+qwen-3.6-35b-a3b-bf16,big-context,05,05-big-context,closed,8,68,6028
+qwen-3.6-35b-a3b-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,31,73974
+qwen-3.6-35b-a3b-bf16,office-baseline,07,07-office-baseline,open,26,26,2316
+qwen-3.6-35b-a3b-bf16,sharegpt,08,08-sharegpt,open,8,8,389
+qwen-3.6-35b-a3b-bf16,reasoning,09,09-reasoning,open,4,4,698
+qwen-3.6-35b-a3b-bf16,monday-peak,10,10-monday-peak,open,37,37,1849
+gemma-4-31b-bf16,chat,01,01-chat,closed,3,18,6385
+gemma-4-31b-bf16,rag-8k,02,02-rag-8k,closed,3,18,45869
+gemma-4-31b-bf16,long-output,03,03-long-output,closed,4,20,2447
+gemma-4-31b-bf16,multi-turn,04,04-multi-turn,closed,3,22,12127
+gemma-4-31b-bf16,big-context,05,05-big-context,closed,3,23,23792
+gemma-4-31b-bf16,concurrency-stress,06,06-concurrency-stress,closed,1,6,367913
+gemma-4-31b-bf16,office-baseline,07,07-office-baseline,open,3,3,248552
+gemma-4-31b-bf16,sharegpt,08,08-sharegpt,open,3,3,1050
+gemma-4-31b-bf16,reasoning,09,09-reasoning,open,2,2,2091
+gemma-4-31b-bf16,monday-peak,10,10-monday-peak,open,13,13,5630
+qwen-3.5-2b-bf16,chat,01,01-chat,closed,50,321,435
+qwen-3.5-2b-bf16,rag-8k,02,02-rag-8k,closed,38,286,2738
+qwen-3.5-2b-bf16,long-output,03,03-long-output,closed,52,245,199
+qwen-3.5-2b-bf16,multi-turn,04,04-multi-turn,closed,48,379,712
+qwen-3.5-2b-bf16,big-context,05,05-big-context,closed,42,332,1359
+qwen-3.5-2b-bf16,concurrency-stress,06,06-concurrency-stress,closed,14,118,17847
+qwen-3.5-2b-bf16,office-baseline,07,07-office-baseline,open,135,135,371
+qwen-3.5-2b-bf16,sharegpt,08,08-sharegpt,open,19,19,73
+qwen-3.5-2b-bf16,reasoning,09,09-reasoning,open,28,28,142
+qwen-3.5-2b-bf16,monday-peak,10,10-monday-peak,open,184,184,432
+ministral-3-3b-instruct-bf16,chat,01,01-chat,closed,42,357,864
+ministral-3-3b-instruct-bf16,rag-8k,02,02-rag-8k,closed,17,158,3403
+ministral-3-3b-instruct-bf16,long-output,03,03-long-output,closed,50,239,461
+ministral-3-3b-instruct-bf16,multi-turn,04,04-multi-turn,closed,35,329,1210
+ministral-3-3b-instruct-bf16,big-context,05,05-big-context,closed,25,224,1704
+ministral-3-3b-instruct-bf16,concurrency-stress,06,06-concurrency-stress,closed,3,44,31187
+ministral-3-3b-instruct-bf16,office-baseline,07,07-office-baseline,open,95,95,660
 ministral-3-3b-instruct-bf16,sharegpt,08,08-sharegpt,open,18,18,69
+ministral-3-3b-instruct-bf16,reasoning,09,09-reasoning,open,11,11,423
+ministral-3-3b-instruct-bf16,monday-peak,10,10-monday-peak,open,80,80,803
+ministral-3-8b-instruct-bf16,chat,01,01-chat,closed,24,184,1248
+ministral-3-8b-instruct-bf16,rag-8k,02,02-rag-8k,closed,15,124,6335
+ministral-3-8b-instruct-bf16,long-output,03,03-long-output,closed,25,118,647
+ministral-3-8b-instruct-bf16,multi-turn,04,04-multi-turn,closed,22,206,1966
+ministral-3-8b-instruct-bf16,big-context,05,05-big-context,closed,18,152,3191
+ministral-3-8b-instruct-bf16,concurrency-stress,06,06-concurrency-stress,closed,3,41,45792
+ministral-3-8b-instruct-bf16,office-baseline,07,07-office-baseline,open,72,72,1287
+ministral-3-8b-instruct-bf16,sharegpt,08,08-sharegpt,open,13,13,136
+ministral-3-8b-instruct-bf16,reasoning,09,09-reasoning,open,9,9,592
+ministral-3-8b-instruct-bf16,monday-peak,10,10-monday-peak,open,69,69,1219
+gemma-4-bf16,chat,01,01-chat,closed,11,91,1555
+gemma-4-bf16,rag-8k,02,02-rag-8k,closed,9,76,8795
+gemma-4-bf16,long-output,03,03-long-output,closed,11,81,511
+gemma-4-bf16,multi-turn,04,04-multi-turn,closed,10,89,2316
+gemma-4-bf16,big-context,05,05-big-context,closed,9,79,4257
+gemma-4-bf16,concurrency-stress,06,06-concurrency-stress,closed,3,29,76760
+gemma-4-bf16,office-baseline,07,07-office-baseline,open,30,30,1433
+gemma-4-bf16,sharegpt,08,08-sharegpt,open,8,8,354
+gemma-4-bf16,reasoning,09,09-reasoning,open,5,5,570
+gemma-4-bf16,monday-peak,10,10-monday-peak,open,41,41,1235
+gemma-4-nvfp4,chat,01,01-chat,closed,21,141,1278
+gemma-4-nvfp4,rag-8k,02,02-rag-8k,closed,16,123,8173
+gemma-4-nvfp4,long-output,03,03-long-output,closed,23,161,382
+gemma-4-nvfp4,multi-turn,04,04-multi-turn,closed,20,172,2017
+gemma-4-nvfp4,big-context,05,05-big-context,closed,17,138,3854
+gemma-4-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,4,33,73575
+gemma-4-nvfp4,office-baseline,07,07-office-baseline,open,82,82,1057
+gemma-4-nvfp4,sharegpt,08,08-sharegpt,open,13,13,154
+gemma-4-nvfp4,reasoning,09,09-reasoning,open,9,9,339
+gemma-4-nvfp4,monday-peak,10,10-monday-peak,open,69,69,958
+gpt-oss-20b-mxfp4,chat,01,01-chat,closed,24,180,878
+gpt-oss-20b-mxfp4,rag-8k,02,02-rag-8k,closed,22,162,6113
+gpt-oss-20b-mxfp4,long-output,03,03-long-output,closed,24,204,287
+gpt-oss-20b-mxfp4,multi-turn,04,04-multi-turn,closed,25,141,1637
+gpt-oss-20b-mxfp4,big-context,05,05-big-context,closed,24,140,2943
+gpt-oss-20b-mxfp4,concurrency-stress,06,06-concurrency-stress,closed,6,48,50054
+gpt-oss-20b-mxfp4,office-baseline,07,07-office-baseline,open,94,94,765
+gpt-oss-20b-mxfp4,sharegpt,08,08-sharegpt,open,15,15,120
+gpt-oss-20b-mxfp4,reasoning,09,09-reasoning,open,13,13,256
+gpt-oss-20b-mxfp4,monday-peak,10,10-monday-peak,open,99,99,749
+qwen-3.5-9b-bf16,chat,01,01-chat,closed,13,123,1823
+qwen-3.5-9b-bf16,rag-8k,02,02-rag-8k,closed,10,84,11334
+qwen-3.5-9b-bf16,long-output,03,03-long-output,closed,13,123,637
+qwen-3.5-9b-bf16,multi-turn,04,04-multi-turn,closed,12,113,3131
+qwen-3.5-9b-bf16,big-context,05,05-big-context,closed,11,86,5938
+qwen-3.5-9b-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,32,71438
+qwen-3.5-9b-bf16,office-baseline,07,07-office-baseline,open,38,38,1865
+qwen-3.5-9b-bf16,sharegpt,08,08-sharegpt,open,9,9,251
+qwen-3.5-9b-bf16,reasoning,09,09-reasoning,open,6,6,531
+qwen-3.5-9b-bf16,monday-peak,10,10-monday-peak,open,49,49,1651
+gemma-4-31b-it-nvfp4,chat,01,01-chat,closed,6,37,8404
+gemma-4-31b-it-nvfp4,rag-8k,02,02-rag-8k,closed,4,26,57922
+gemma-4-31b-it-nvfp4,long-output,03,03-long-output,closed,7,36,2707
+gemma-4-31b-it-nvfp4,multi-turn,04,04-multi-turn,closed,6,43,16586
+gemma-4-31b-it-nvfp4,big-context,05,05-big-context,closed,5,32,29898
+gemma-4-31b-it-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,1,6,419863
+gemma-4-31b-it-nvfp4,office-baseline,07,07-office-baseline,open,3,3,219174
+gemma-4-31b-it-nvfp4,sharegpt,08,08-sharegpt,open,6,6,671
+gemma-4-31b-it-nvfp4,reasoning,09,09-reasoning,open,3,3,2147
+gemma-4-31b-it-nvfp4,monday-peak,10,10-monday-peak,open,17,17,5964
+gemma-4-e2b-it-bf16,chat,01,01-chat,closed,43,254,515
+gemma-4-e2b-it-bf16,rag-8k,02,02-rag-8k,closed,33,261,4320
+gemma-4-e2b-it-bf16,long-output,03,03-long-output,closed,45,268,200
+gemma-4-e2b-it-bf16,multi-turn,04,04-multi-turn,closed,42,375,1005
+gemma-4-e2b-it-bf16,big-context,05,05-big-context,closed,36,288,1974
+gemma-4-e2b-it-bf16,concurrency-stress,06,06-concurrency-stress,closed,8,59,42854
+gemma-4-e2b-it-bf16,office-baseline,07,07-office-baseline,open,122,122,493
+gemma-4-e2b-it-bf16,sharegpt,08,08-sharegpt,open,19,19,90
+gemma-4-e2b-it-bf16,reasoning,09,09-reasoning,open,25,25,187
+gemma-4-e2b-it-bf16,monday-peak,10,10-monday-peak,open,154,154,504
+gemma-4-e4b-it-bf16,chat,01,01-chat,closed,21,134,1051
+gemma-4-e4b-it-bf16,rag-8k,02,02-rag-8k,closed,16,118,6935
+gemma-4-e4b-it-bf16,long-output,03,03-long-output,closed,22,126,403
+gemma-4-e4b-it-bf16,multi-turn,04,04-multi-turn,closed,21,145,1806
+gemma-4-e4b-it-bf16,big-context,05,05-big-context,closed,18,136,3371
+gemma-4-e4b-it-bf16,concurrency-stress,06,06-concurrency-stress,closed,5,41,55760
+gemma-4-e4b-it-bf16,office-baseline,07,07-office-baseline,open,82,82,889
+gemma-4-e4b-it-bf16,sharegpt,08,08-sharegpt,open,13,13,165
+gemma-4-e4b-it-bf16,reasoning,09,09-reasoning,open,10,10,318
+gemma-4-e4b-it-bf16,monday-peak,10,10-monday-peak,open,86,86,809
+granite-4-1-8b-bf16,chat,01,01-chat,closed,12,60,1887
+granite-4-1-8b-bf16,rag-8k,02,02-rag-8k,closed,9,69,10825
+granite-4-1-8b-bf16,long-output,03,03-long-output,closed,13,76,710
+granite-4-1-8b-bf16,multi-turn,04,04-multi-turn,closed,12,100,3158
+granite-4-1-8b-bf16,big-context,05,05-big-context,closed,10,84,5566
+granite-4-1-8b-bf16,concurrency-stress,06,06-concurrency-stress,closed,2,26,74995
+granite-4-1-8b-bf16,office-baseline,07,07-office-baseline,open,31,31,1751
+granite-4-1-8b-bf16,sharegpt,08,08-sharegpt,open,9,9,255
+granite-4-1-8b-bf16,reasoning,09,09-reasoning,open,5,5,523
+granite-4-1-8b-bf16,monday-peak,10,10-monday-peak,open,44,44,1490
+kat-coder-v2-5-bf16,chat,01,01-chat,closed,10,76,2010
+kat-coder-v2-5-bf16,rag-8k,02,02-rag-8k,closed,8,66,9445
+kat-coder-v2-5-bf16,long-output,03,03-long-output,closed,11,65,739
+kat-coder-v2-5-bf16,multi-turn,04,04-multi-turn,closed,9,78,3071
+kat-coder-v2-5-bf16,big-context,05,05-big-context,closed,9,75,5104
+kat-coder-v2-5-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,38,54142
+kat-coder-v2-5-bf16,office-baseline,07,07-office-baseline,open,26,26,1617
+kat-coder-v2-5-bf16,sharegpt,08,08-sharegpt,open,8,8,386
+kat-coder-v2-5-bf16,reasoning,09,09-reasoning,open,4,4,712
+kat-coder-v2-5-bf16,monday-peak,10,10-monday-peak,open,37,37,1366
+lfm2-5-2-6b-bf16,chat,01,01-chat,closed,38,370,750
+lfm2-5-2-6b-bf16,rag-8k,02,02-rag-8k,closed,30,255,3345
+lfm2-5-2-6b-bf16,long-output,03,03-long-output,closed,39,243,214
+lfm2-5-2-6b-bf16,multi-turn,04,04-multi-turn,closed,37,345,1044
+lfm2-5-2-6b-bf16,big-context,05,05-big-context,closed,32,266,1764
+lfm2-5-2-6b-bf16,concurrency-stress,06,06-concurrency-stress,closed,10,102,21230
+lfm2-5-2-6b-bf16,office-baseline,07,07-office-baseline,open,122,122,467
+lfm2-5-2-6b-bf16,sharegpt,08,08-sharegpt,open,19,19,86
+lfm2-5-2-6b-bf16,reasoning,09,09-reasoning,open,22,22,174
+lfm2-5-2-6b-bf16,monday-peak,10,10-monday-peak,open,149,149,454
+mistral-small-4-119b-nvfp4,chat,01,01-chat,closed,12,80,2136
+mistral-small-4-119b-nvfp4,rag-8k,02,02-rag-8k,closed,9,61,11125
+mistral-small-4-119b-nvfp4,long-output,03,03-long-output,closed,12,83,697
+mistral-small-4-119b-nvfp4,multi-turn,04,04-multi-turn,closed,10,89,3512
+mistral-small-4-119b-nvfp4,big-context,05,05-big-context,closed,9,72,5946
+mistral-small-4-119b-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,3,22,73818
+mistral-small-4-119b-nvfp4,office-baseline,07,07-office-baseline,open,21,21,1984
+mistral-small-4-119b-nvfp4,sharegpt,08,08-sharegpt,open,10,10,402
+mistral-small-4-119b-nvfp4,reasoning,09,09-reasoning,open,3,3,763
+mistral-small-4-119b-nvfp4,monday-peak,10,10-monday-peak,open,35,35,1549
+muse-glimmer-30b-bf16,chat,01,01-chat,closed,4,41,4927
+muse-glimmer-30b-bf16,rag-8k,02,02-rag-8k,closed,4,32,28816
+muse-glimmer-30b-bf16,long-output,03,03-long-output,closed,4,34,1598
+muse-glimmer-30b-bf16,multi-turn,04,04-multi-turn,closed,4,40,8327
+muse-glimmer-30b-bf16,big-context,05,05-big-context,closed,4,36,15501
+muse-glimmer-30b-bf16,concurrency-stress,06,06-concurrency-stress,closed,2,14,167973
+muse-glimmer-30b-bf16,office-baseline,07,07-office-baseline,open,3,3,17039
+muse-glimmer-30b-bf16,sharegpt,08,08-sharegpt,open,5,5,763
+muse-glimmer-30b-bf16,reasoning,09,09-reasoning,open,2,2,1507
+muse-glimmer-30b-bf16,monday-peak,10,10-monday-peak,open,18,18,3842
+muse-glimmer-30b-bf16-spec,chat,01,01-chat,closed,13,86,5667
+muse-glimmer-30b-bf16-spec,rag-8k,02,02-rag-8k,closed,7,44,29450
+muse-glimmer-30b-bf16-spec,long-output,03,03-long-output,closed,12,90,1896
+muse-glimmer-30b-bf16-spec,multi-turn,04,04-multi-turn,closed,13,64,9079
+muse-glimmer-30b-bf16-spec,big-context,05,05-big-context,closed,7,49,15699
+muse-glimmer-30b-bf16-spec,concurrency-stress,06,06-concurrency-stress,closed,2,14,171624
+muse-glimmer-30b-bf16-spec,office-baseline,07,07-office-baseline,open,2,2,59120
+muse-glimmer-30b-bf16-spec,sharegpt,08,08-sharegpt,open,10,10,941
+muse-glimmer-30b-bf16-spec,reasoning,09,09-reasoning,open,3,3,2602
+muse-glimmer-30b-bf16-spec,monday-peak,10,10-monday-peak,open,13,13,4857
+nemotron-3-nano-4b-fp8,chat,01,01-chat,closed,35,308,931
+nemotron-3-nano-4b-fp8,rag-8k,02,02-rag-8k,closed,28,225,3853
+nemotron-3-nano-4b-fp8,long-output,03,03-long-output,closed,37,197,331
+nemotron-3-nano-4b-fp8,multi-turn,04,04-multi-turn,closed,33,300,1246
+nemotron-3-nano-4b-fp8,big-context,05,05-big-context,closed,29,243,1918
+nemotron-3-nano-4b-fp8,concurrency-stress,06,06-concurrency-stress,closed,10,91,24235
+nemotron-3-nano-4b-fp8,office-baseline,07,07-office-baseline,open,102,102,733
+nemotron-3-nano-4b-fp8,sharegpt,08,08-sharegpt,open,17,17,107
+nemotron-3-nano-4b-fp8,reasoning,09,09-reasoning,open,15,15,467
+nemotron-3-nano-4b-fp8,monday-peak,10,10-monday-peak,open,113,113,774
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,chat,01,01-chat,closed,9,73,1477
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,rag-8k,02,02-rag-8k,closed,8,67,6258
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,long-output,03,03-long-output,closed,10,63,688
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,multi-turn,04,04-multi-turn,closed,8,74,2130
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,big-context,05,05-big-context,closed,8,69,3460
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,45,36919
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,office-baseline,07,07-office-baseline,open,29,29,1214
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,sharegpt,08,08-sharegpt,open,8,8,433
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,reasoning,09,09-reasoning,open,5,5,650
+nemotron-3-nano-omni-30b-a3b-reasoning-bf16,monday-peak,10,10-monday-peak,open,41,41,1111
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,chat,01,01-chat,closed,17,145,1416
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,rag-8k,02,02-rag-8k,closed,13,117,5156
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,long-output,03,03-long-output,closed,19,121,863
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,multi-turn,04,04-multi-turn,closed,15,137,1998
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,big-context,05,05-big-context,closed,14,122,2871
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,concurrency-stress,06,06-concurrency-stress,closed,6,71,26828
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,office-baseline,07,07-office-baseline,open,54,54,1280
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,sharegpt,08,08-sharegpt,open,13,13,196
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,reasoning,09,09-reasoning,open,8,8,816
+nemotron-3-nano-omni-30b-a3b-reasoning-fp8,monday-peak,10,10-monday-peak,open,65,65,1225
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,chat,01,01-chat,closed,24,219,1475
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,rag-8k,02,02-rag-8k,closed,20,153,5974
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,long-output,03,03-long-output,closed,29,159,821
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,multi-turn,04,04-multi-turn,closed,24,207,2143
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,big-context,05,05-big-context,closed,20,160,3351
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,7,68,32651
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,office-baseline,07,07-office-baseline,open,89,89,1089
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,sharegpt,08,08-sharegpt,open,15,15,130
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,reasoning,09,09-reasoning,open,16,16,588
+nemotron-3-nano-omni-30b-a3b-reasoning-nvfp4,monday-peak,10,10-monday-peak,open,87,87,1005
+nemotron-cascade-2-30b-a3b-bf16,chat,01,01-chat,closed,8,75,1497
+nemotron-cascade-2-30b-a3b-bf16,rag-8k,02,02-rag-8k,closed,7,67,6215
+nemotron-cascade-2-30b-a3b-bf16,long-output,03,03-long-output,closed,9,68,673
+nemotron-cascade-2-30b-a3b-bf16,multi-turn,04,04-multi-turn,closed,8,74,2127
+nemotron-cascade-2-30b-a3b-bf16,big-context,05,05-big-context,closed,8,70,3454
+nemotron-cascade-2-30b-a3b-bf16,concurrency-stress,06,06-concurrency-stress,closed,4,49,35919
+nemotron-cascade-2-30b-a3b-bf16,office-baseline,07,07-office-baseline,open,25,25,1298
+nemotron-cascade-2-30b-a3b-bf16,sharegpt,08,08-sharegpt,open,7,7,489
+nemotron-cascade-2-30b-a3b-bf16,reasoning,09,09-reasoning,open,4,4,706
+nemotron-cascade-2-30b-a3b-bf16,monday-peak,10,10-monday-peak,open,36,36,1140
+qwen-3-5-4b-bf16,chat,01,01-chat,closed,22,216,987
+qwen-3-5-4b-bf16,rag-8k,02,02-rag-8k,closed,18,142,7145
+qwen-3-5-4b-bf16,long-output,03,03-long-output,closed,22,219,389
+qwen-3-5-4b-bf16,multi-turn,04,04-multi-turn,closed,21,199,1732
+qwen-3-5-4b-bf16,big-context,05,05-big-context,closed,18,149,3348
+qwen-3-5-4b-bf16,concurrency-stress,06,06-concurrency-stress,closed,6,51,44648
+qwen-3-5-4b-bf16,office-baseline,07,07-office-baseline,open,82,82,961
+qwen-3-5-4b-bf16,sharegpt,08,08-sharegpt,open,13,13,146
+qwen-3-5-4b-bf16,reasoning,09,09-reasoning,open,9,9,318
+qwen-3-5-4b-bf16,monday-peak,10,10-monday-peak,open,84,84,980
+qwen-3-6-27b-bf16,chat,01,01-chat,closed,4,40,5883
+qwen-3-6-27b-bf16,rag-8k,02,02-rag-8k,closed,3,27,37902
+qwen-3-6-27b-bf16,long-output,03,03-long-output,closed,4,31,2112
+qwen-3-6-27b-bf16,multi-turn,04,04-multi-turn,closed,4,36,10386
+qwen-3-6-27b-bf16,big-context,05,05-big-context,closed,3,28,19417
+qwen-3-6-27b-bf16,concurrency-stress,06,06-concurrency-stress,closed,1,10,231009
+qwen-3-6-27b-bf16,office-baseline,07,07-office-baseline,open,4,4,82171
+qwen-3-6-27b-bf16,sharegpt,08,08-sharegpt,open,3,3,989
+qwen-3-6-27b-bf16,reasoning,09,09-reasoning,open,2,2,1777
+qwen-3-6-27b-bf16,monday-peak,10,10-monday-peak,open,16,16,5422
+qwen-3-6-35b-a3b-nvfp4,chat,01,01-chat,closed,20,196,1214
+qwen-3-6-35b-a3b-nvfp4,rag-8k,02,02-rag-8k,closed,17,135,7729
+qwen-3-6-35b-a3b-nvfp4,long-output,03,03-long-output,closed,22,176,466
+qwen-3-6-35b-a3b-nvfp4,multi-turn,04,04-multi-turn,closed,20,184,2208
+qwen-3-6-35b-a3b-nvfp4,big-context,05,05-big-context,closed,17,138,3995
+qwen-3-6-35b-a3b-nvfp4,concurrency-stress,06,06-concurrency-stress,closed,6,48,50719
+qwen-3-6-35b-a3b-nvfp4,office-baseline,07,07-office-baseline,open,82,82,1083
+qwen-3-6-35b-a3b-nvfp4,sharegpt,08,08-sharegpt,open,15,15,144
+qwen-3-6-35b-a3b-nvfp4,reasoning,09,09-reasoning,open,8,8,380
+qwen-3-6-35b-a3b-nvfp4,monday-peak,10,10-monday-peak,open,72,72,1074
+qwen-3-8-27b-bf16,chat,01,01-chat,closed,4,39,6086
+qwen-3-8-27b-bf16,rag-8k,02,02-rag-8k,closed,3,27,37787
+qwen-3-8-27b-bf16,long-output,03,03-long-output,closed,4,38,2141
+qwen-3-8-27b-bf16,multi-turn,04,04-multi-turn,closed,4,36,10259
+qwen-3-8-27b-bf16,big-context,05,05-big-context,closed,3,28,19383
+qwen-3-8-27b-bf16,concurrency-stress,06,06-concurrency-stress,closed,1,10,228576
+qwen-3-8-27b-bf16,office-baseline,07,07-office-baseline,open,4,4,87195
+qwen-3-8-27b-bf16,sharegpt,08,08-sharegpt,open,4,4,1008
+qwen-3-8-27b-bf16,reasoning,09,09-reasoning,open,2,2,1797
+qwen-3-8-27b-bf16,monday-peak,10,10-monday-peak,open,15,15,5394
+qwen-3-8-27b-fp8,chat,01,01-chat,closed,7,67,10863
+qwen-3-8-27b-fp8,rag-8k,02,02-rag-8k,closed,4,28,70145
+qwen-3-8-27b-fp8,long-output,03,03-long-output,closed,7,66,3688
+qwen-3-8-27b-fp8,multi-turn,04,04-multi-turn,closed,6,53,19286
+qwen-3-8-27b-fp8,big-context,05,05-big-context,closed,5,29,36226
+qwen-3-8-27b-fp8,concurrency-stress,06,06-concurrency-stress,closed,1,7,409504
+qwen-3-8-27b-fp8,office-baseline,07,07-office-baseline,open,3,3,426055
+qwen-3-8-27b-fp8,sharegpt,08,08-sharegpt,open,7,7,908
+qwen-3-8-27b-fp8,reasoning,09,09-reasoning,open,4,4,3069
+qwen-3-8-27b-fp8,monday-peak,10,10-monday-peak,open,16,16,8814