yzeng58 commited on
Commit
2c6f8c8
·
verified ·
1 Parent(s): 2afd3c6

Add BenchPress score matrix dataset

Browse files
README.md ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ task_categories:
4
+ - tabular-regression
5
+ language:
6
+ - en
7
+ pretty_name: BenchPress Score Matrix
8
+ configs:
9
+ - config_name: scores_all
10
+ data_files:
11
+ - split: train
12
+ path: data/scores_all.parquet
13
+ - config_name: scores_paper
14
+ data_files:
15
+ - split: train
16
+ path: data/scores_paper.parquet
17
+ - config_name: models
18
+ data_files:
19
+ - split: train
20
+ path: data/models.parquet
21
+ - config_name: benchmarks
22
+ data_files:
23
+ - split: train
24
+ path: data/benchmarks.parquet
25
+ ---
26
+
27
+ # BenchPress Score Matrix
28
+
29
+ This dataset contains the public model-by-benchmark score matrix used by
30
+ BenchPress. The release is a tabular artifact: model metadata, benchmark
31
+ metadata, one row per observed score, and the paper-canonical dense subset used
32
+ in the BenchPress experiments.
33
+
34
+ The source repository is
35
+ [`anadim/BenchPress`](https://github.com/anadim/BenchPress). This export was
36
+ generated from commit `67d07b70a5fc5c1bca666f85b235502bb6b6577b`. The export came from a working tree with local data changes.
37
+
38
+ ## Files
39
+
40
+ | File | Contents |
41
+ |---|---|
42
+ | `data/scores_all.csv` / `.parquet` | All numeric score rows in the audit pool, with source and audit metadata. |
43
+ | `data/scores_paper.csv` / `.parquet` | Long-form rows for the paper-canonical matrix. |
44
+ | `data/models.csv` / `.parquet` | Model metadata and canonical evaluation settings. |
45
+ | `data/benchmarks.csv` / `.parquet` | Benchmark metadata and canonical benchmark settings. |
46
+ | `data/score_matrix_paper_wide.csv` | Wide model × benchmark matrix for the paper-canonical subset. |
47
+ | `metadata.json` | Export counts, source commit, and matrix construction metadata. |
48
+
49
+ ## Quick start
50
+
51
+ ```python
52
+ from datasets import load_dataset
53
+
54
+ scores = load_dataset("yzeng58/benchpress-score-matrix", "scores_paper")["train"].to_pandas()
55
+ models = load_dataset("yzeng58/benchpress-score-matrix", "models")["train"].to_pandas()
56
+ benchmarks = load_dataset("yzeng58/benchpress-score-matrix", "benchmarks")["train"].to_pandas()
57
+ ```
58
+
59
+ For a complete audit-pool table:
60
+
61
+ ```python
62
+ scores_all = load_dataset("yzeng58/benchpress-score-matrix", "scores_all")["train"].to_pandas()
63
+ ```
64
+
65
+ ## Schema
66
+
67
+ `scores_all` and `scores_paper` include:
68
+
69
+ - `model_id`, `model_name`, `provider`
70
+ - `benchmark_id`, `benchmark_name`, `category`, `metric`
71
+ - `score`
72
+ - `reference_url`, `source_type`, `audit_status`, `matches_canonical`
73
+ - `reported_setting_json`, `notes`
74
+
75
+ `models` and `benchmarks` include an `in_paper_matrix` flag that identifies
76
+ rows retained by the paper-canonical threshold filter.
77
+
78
+ ## Matrix construction
79
+
80
+ The paper-canonical matrix applies the BenchPress construction pipeline:
81
+ audit-status filtering, canonical representative selection, and the iterative
82
+ threshold filter. Current export counts:
83
+
84
+ - audit pool: 189 models, 316 benchmarks, 4903 numeric score rows
85
+ - paper matrix: 84 models × 133 benchmarks, 2604 observed cells (23.3% fill)
86
+
87
+ ## Caveats
88
+
89
+ Scores come from heterogeneous public sources: model cards, official blogs,
90
+ technical reports, benchmark leaderboards, and third-party aggregators. Each row
91
+ keeps the source URL, source type, audit status, and canonical-setting match
92
+ flag so downstream users can choose their own filtering policy.
data/benchmarks.csv ADDED
The diff for this file is too large to render. See raw diff
 
data/benchmarks.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa1d5e1c616c86c6db29a8ebe8274f153e879a3d9a32ad2431bfe06860d2f931
3
+ size 37708
data/models.csv ADDED
@@ -0,0 +1,190 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model_id,model_name,provider,release_date,params_total_M,params_active_M,architecture,is_reasoning,open_weights,canonical_setting_json,in_paper_matrix
2
+ qwen-2.5-1.5b,Qwen2.5-1.5B-Instruct,Alibaba,2024-09-19,1500.0,1500.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
3
+ qwen-2.5-14b,Qwen2.5-14B-Instruct,Alibaba,2024-09-19,14000.0,14000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
4
+ qwen-2.5-14b-instruct,Qwen2.5-14B,Alibaba,2024-09-19,14000.0,14000.0,dense,False,True,"{""context"": ""default (128K)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Phi-4 paper Table 1 baseline."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
5
+ qwen-2.5-32b,Qwen2.5-32B-Instruct,Alibaba,2024-09-19,32000.0,32000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
6
+ qwen-2.5-7b,Qwen2.5-7B-Instruct,Alibaba,2024-09-19,7000.0,7000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
7
+ qwen2-72b-instruct,Qwen2 72B Instruct,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
8
+ qwen2-7b-instruct,Qwen2 7B Instruct,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
9
+ qwen2.5-72b-instruct,Qwen2.5 72B Instruct,Alibaba,2024-09-19,,,,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
10
+ qwen2.5-coder-32b-instruct,Qwen2.5-Coder 32B Instruct,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
11
+ qwen2.5-coder-7b-instruct,Qwen2.5-Coder 7B Instruct,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
12
+ qwen2.5-omni-7b,Qwen2.5-Omni-7B,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
13
+ qwen2.5-vl-32b-instruct,Qwen2.5 VL 32B Instruct,Alibaba,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
14
+ qwen3-0.6b,Qwen3-0.6B,Alibaba,2025-05-15,600.0,600.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
15
+ qwen3-1.7b,Qwen3-1.7B,Alibaba,2025-05-15,1700.0,1700.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
16
+ qwen3-14b,Qwen3-14B,Alibaba,2025-05-15,14000.0,14000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
17
+ qwen3-235b,Qwen3-235B-A22B,Alibaba,2025-05-15,235000.0,22000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 11 Thinking column. Output length 32768 (38912 for AIME)."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
18
+ qwen3-30b-a3b,Qwen3-30B-A3B,Alibaba,2025-05-15,30000.0,3000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
19
+ qwen3-32b,Qwen3-32B,Alibaba,2025-05-15,32000.0,32000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
20
+ qwen3-4b,Qwen3-4B,Alibaba,2025-05-15,4000.0,4000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
21
+ qwen3-8b,Qwen3-8B,Alibaba,2025-05-15,8000.0,8000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (Qwen-Agent / sglang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3 tech report (arxiv:2505.09388) Table 13/15/17/19 Thinking column."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
22
+ qwen3.5-397b,Qwen3.5-397B-A17B,Alibaba,2026-02-01,397000.0,17000.0,MoE,True,True,"{""context"": ""default (262K)"", ""effort"": ""default (96k thinking budget)"", ""harness"": ""official (sglang/vllm with --reasoning-parser qwen3)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Qwen3.5 HF model card. Vision-language MoE 397B/17B."", ""prompt_style"": ""default (thinking by default)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
23
+ qwen3.6-plus,Qwen3.6-Plus,Alibaba,2026-03-31,,,,True,False,,True
24
+ qwq-32b,QwQ-32B,Alibaba,2025-03-05,32800.0,32800.0,Dense,True,True,"{""context"": ""default (32K)"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per QwQ-32B release blog (qwenlm.github.io/blog/qwq-32b/) + Qwen3 tech report Table 13 baseline column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
25
+ olmo-1b,OLMo 1B 0724,Allen AI,2024-07-01,1000.0,1000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
26
+ olmo-2-13b,OLMo 2 13B,Allen AI,2025-01-01,13000.0,13000.0,Dense,False,True,,False
27
+ olmo-2-1b,OLMo 2 1B Instruct,Allen AI,2025-01-03,1000.0,1000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
28
+ olmo-2-32b,OLMo 2 32B Instruct,Allen AI,2025-01-03,32000.0,32000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
29
+ olmo-2-7b,OLMo 2 7B Instruct,Allen AI,2025-01-03,7000.0,7000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
30
+ olmo-7b-0424,OLMo 7B 0424 Instruct,Allen AI,2024-04-17,7000.0,7000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
31
+ tulu-3-8b,Tulu 3 8B,Allen AI,2024-11-22,8000.0,8000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
32
+ amazon-nova-premier,Amazon Nova Premier,Amazon,2025-04-01,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Amazon Nova Premier tech report (Apr 2025). SimpleQA cell reported with SerpApi web tool — see cell-level reported_setting."", ""prompt_style"": ""zeroshot-cot"", ""sampling"": ""pass@1"", ""temperature"": 0.0, ""tools"": ""none""}",False
33
+ amazon-nova-pro,Amazon Nova Pro,Amazon,2025-01-01,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Amazon Nova Premier tech report (cross-reports Nova Pro) and Nova family arxiv 2506.12103. SimpleQA cell reported with SerpApi web tool — see cell-level reported_setting."", ""prompt_style"": ""zeroshot-cot"", ""sampling"": ""pass@1"", ""temperature"": 0.0, ""tools"": ""none""}",False
34
+ nova-lite,Nova Lite,Amazon,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
35
+ nova-micro,Nova Micro,Amazon,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
36
+ claude-3-haiku,Claude 3 Haiku,Anthropic,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
37
+ claude-3-opus,Claude 3 Opus,Anthropic,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
38
+ claude-3-sonnet,Claude 3 Sonnet,Anthropic,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
39
+ claude-3.5-haiku,Claude 3.5 Haiku,Anthropic,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
40
+ claude-3.5-sonnet,Claude 3.5 Sonnet (1022),Anthropic,2024-10-22,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
41
+ claude-3.7-sonnet,Claude 3.7 Sonnet,Anthropic,2025-02-24,,,,True,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Claude 4 blog cross-model table: with extended thinking for GPQA/MMMU/AIME/TAU; no extended thinking for SWE/Terminal/MMMLU."", ""prompt_style"": ""zeroshot-cot"", ""sampling"": ""pass@1"", ""temperature"": 0.0, ""tools"": ""none""}",True
42
+ claude-haiku-4.5,Claude Haiku 4.5,Anthropic,2025-10-15,,,,True,False,"{""effort"": ""128K thinking budget"", ""harness"": ""official"", ""judge"": ""official harness"", ""mode"": ""thinking"", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg over ≥10 runs unless noted)"", ""temperature"": ""default (Anthropic default)"", ""tools"": ""none (per-bench override for agentic benchmarks)""}",True
43
+ claude-mythos,Claude Mythos Preview,Anthropic,2026-04-07,,,,True,False,"{""context"": ""up to 1M"", ""effort"": ""adaptive thinking at max effort"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Mythos System Card Table 6.3.A: standard config is adaptive thinking max effort, default sampling (temperature, top_p), avg 5 trials, context up to 1M."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg 5 trials)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
44
+ claude-opus-4,Claude Opus 4,Anthropic,2025-05-22,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""Claude Code agent framework (for agentic benchmarks)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Claude 4 announcement table: single-pass values; no extended thinking for SWE/Terminal/MMMLU; extended thinking up to 64K for GPQA/MMMU/AIME/TAU; tools=agentic for SWE/Terminal/TAU/OSWorld. Agentic harness=Claude Code per Claude 4 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single-pass)"", ""temperature"": ""top_p=0.95"", ""tools"": ""none (per-bench override)""}",True
45
+ claude-opus-4.1,Claude Opus 4.1,Anthropic,2025-08-05,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""Terminus 1 (for agentic benchmarks)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Opus 4.1 announcement table. Single-pass values; harness=Terminus 1 for Terminal-Bench (note: differs from Claude 4 blog which used Claude Code framework)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single-pass)"", ""temperature"": ""top_p=0.95"", ""tools"": ""none (per-bench override)""}",True
46
+ claude-opus-4.5,Claude Opus 4.5,Anthropic,2025-11-24,,,,True,False,"{""context"": ""200K (default)"", ""effort"": ""high (default)"", ""harness"": ""Terminus-2 (for terminal-bench); official otherwise"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Opus 4.5 announcement: 64K thinking budget, interleaved scratchpads, 200K context, default effort=high, default sampling, avg 5 trials. Exceptions: SWE-bench Verified=no thinking; Terminal-Bench=128K thinking budget."", ""prompt_style"": ""interleaved scratchpads"", ""sampling"": ""pass@1 (avg 5 trials)"", ""temperature"": ""default (top_p, temperature)"", ""tools"": ""none (per-bench override)""}",True
47
+ claude-opus-4.6,Claude Opus 4.6,Anthropic,2026-02-05,,,,True,False,"{""context"": ""up to 1M"", ""effort"": ""max"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Mythos System Card Table 6.3.A: standard config = adaptive thinking max effort, default sampling, avg 5 trials, context up to 1M."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg 5 trials)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
48
+ claude-opus-4.7,Claude Opus 4.7,Anthropic,2026-04-22,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""Terminus-2 (for terminal-bench, thinking disabled)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Opus 4.7 announcement table. Default thinking mode with extended thinking; Terminal-Bench uses Terminus-2 harness with thinking disabled (1× guaranteed/3× ceiling, avg 5 attempts)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg 5 trials)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
49
+ claude-sonnet-3.7,Claude Sonnet 3.7,Anthropic,2025-02-24,,,,False,False,,False
50
+ claude-sonnet-4,Claude Sonnet 4,Anthropic,2025-05-22,,,,False,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""Claude Code agent framework (for agentic benchmarks)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Claude 4 announcement table: single-pass values; no extended thinking for SWE/Terminal/MMMLU; extended thinking up to 64K for GPQA/MMMU/AIME/TAU; tools=agentic for SWE/Terminal/TAU/OSWorld. Agentic harness=Claude Code per Claude 4 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single-pass)"", ""temperature"": ""top_p=0.95"", ""tools"": ""none (per-bench override)""}",True
51
+ claude-sonnet-4.5,Claude Sonnet 4.5,Anthropic,2025-09-29,,,,True,False,"{""context"": ""200K (default)"", ""effort"": ""high (default)"", ""harness"": ""Terminus-2 (for terminal-bench); official otherwise"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Opus 4.5 announcement: 64K thinking budget, interleaved scratchpads, 200K context, default effort=high, default sampling, avg 5 trials. Exceptions: SWE-bench Verified=no thinking; Terminal-Bench=128K thinking budget."", ""prompt_style"": ""interleaved scratchpads"", ""sampling"": ""pass@1 (avg 5 trials)"", ""temperature"": ""default (top_p, temperature)"", ""tools"": ""none (per-bench override)""}",True
52
+ claude-sonnet-4.6,Claude Sonnet 4.6,Anthropic,2026-02-17,,,,True,False,"{""context"": ""default"", ""effort"": ""max"", ""harness"": ""Terminus-2 (terminal-bench, thinking off); official otherwise"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Anthropic Sonnet 4.6 announcement: max thinking effort default; Terminal-Bench=thinking off, Terminus-2; SWE-bench=avg 10 trials; BrowseComp/HLE-with-tools have specific tool configs (web search/fetch + 50k context compaction)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg 5-15 trials)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
53
+ doubao-seed-2.0-pro,Doubao Seed 2.0 Pro,ByteDance,2026-02-14,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Doubao Seed 2.0 Pro model card. High thinking config."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
54
+ seed-thinking-v1.5,Seed-Thinking-v1.5,ByteDance,2025-04-01,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Seed-Thinking-v1.5 paper."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
55
+ command-a,Command A,Cohere,2025-03-13,111000.0,,Dense,False,True,"{""context"": ""default (256K)"", ""effort"": ""n/a"", ""harness"": ""simple-evals (for MMLU/MMLU-Pro/GPQA); internal reproductions for code/SQL"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Command A tech report (cohere.com/research/papers/command-a-technical-report.pdf)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
56
+ command-r7b,Command R7B,Cohere,2024-12-13,7000.0,7000.0,dense,False,True,"{""context"": ""128K"", ""effort"": ""n/a"", ""harness"": ""simple-evals"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Command A tech report (Cohere first-party)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
57
+ deepseek-3.1,DeepSeek 3.1,DeepSeek,2025-08-21,685000.0,37000.0,MoE,False,True,,False
58
+ deepseek-r1,DeepSeek-R1,DeepSeek,2025-01-20,671000.0,37000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1-0528 model card."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
59
+ deepseek-r1-0528,DeepSeek-R1-0528,DeepSeek,2025-05-28,671000.0,37000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1-0528 model card."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
60
+ deepseek-r1-distill-llama-70b,DeepSeek-R1-Distill-Llama-70B,DeepSeek,2025-01-20,70000.0,70000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official (DeepSeek SGLang/vLLM)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Distill of DeepSeek-R1 onto Llama-70B base. Cells in BP currently primarily sourced from Qwen3 tech report Table 13 third-party self-test (matches_canonical=false). DeepSeek own paper audit pending — canonical_setting may need refinement."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
61
+ deepseek-r1-distill-llama-8b,DeepSeek-R1-Distill-Llama-8B,DeepSeek,2025-01-20,8000.0,8000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1 paper Table 15 (distilled model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
62
+ deepseek-r1-distill-qwen-1.5b,DeepSeek-R1-Distill-Qwen-1.5B,DeepSeek,2025-01-20,1500.0,1500.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1 paper Table 15 (distilled model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
63
+ deepseek-r1-distill-qwen-14b,DeepSeek-R1-Distill-Qwen-14B,DeepSeek,2025-01-20,14000.0,14000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1 paper Table 15 (distilled model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
64
+ deepseek-r1-distill-qwen-32b,DeepSeek-R1-Distill-Qwen-32B,DeepSeek,2025-01-20,32000.0,32000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1 paper Table 15 (distilled model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
65
+ deepseek-r1-distill-qwen-7b,DeepSeek-R1-Distill-Qwen-7B,DeepSeek,2025-01-20,7000.0,7000.0,Dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek R1 paper Table 15 (distilled model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
66
+ deepseek-v2-0506,DeepSeek-V2-0506,DeepSeek,2024-05-06,,,,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
67
+ deepseek-v2.5-0905,DeepSeek-V2.5-0905,DeepSeek,2024-09-05,,,,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
68
+ deepseek-v3,DeepSeek-V3,DeepSeek,2025-01-01,671000.0,37000.0,MoE,False,True,,True
69
+ deepseek-v3-0324,DeepSeek-V3-0324,DeepSeek,2025-03-24,671000.0,37000.0,MoE,False,True,,False
70
+ deepseek-v3.2,DeepSeek-V3.2,DeepSeek,2025-12-01,671000.0,37000.0,MoE,True,True,"{""context"": ""128K"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek V3.2 tech report: temperature=1.0, context=128K, thinking mode for tool-use."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""1.0"", ""tools"": ""none (per-bench override)""}",True
71
+ deepseek-v3.2-speciale,DeepSeek-V3.2-Speciale,DeepSeek,2025-12-01,671000.0,37000.0,MoE,True,True,"{""context"": ""128K"", ""effort"": ""max"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""DS V3.2-Speciale: high-compute reasoning variant of V3.2. Per V3.2 tech report Table 3."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""1.0"", ""tools"": ""none (per-bench override)""}",False
72
+ deepseek-v4-flash,DeepSeek-V4-Flash,DeepSeek,2026-04-22,284000.0,13000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""Max"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek V4-Pro model card. Three reasoning modes available (Non-Think/High/Max); BP canonical = Max mode (most powerful). Other modes' values stored as candidates if needed."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
73
+ deepseek-v4-pro,DeepSeek-V4-Pro,DeepSeek,2026-04-22,1600000.0,49000.0,MoE,True,True,"{""context"": ""default"", ""effort"": ""Max"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per DeepSeek V4-Pro model card. Three reasoning modes available (Non-Think/High/Max); BP canonical = Max mode (most powerful). Other modes' values stored as candidates if needed."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
74
+ gemini-1.5-flash,Gemini 1.5 Flash,Google,2024-05-14,,,moe,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemini 2.0 Flash Model Card (April 15 2025): non-thinking generation, default sampling. Same canonical setting applies to 1.5 / 2.0 family."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
75
+ gemini-1.5-flash-8b,Gemini 1.5 Flash 8B,Google,2024-10-03,8000.0,8000.0,dense,False,False,"{""context"": ""1M"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Command A paper baseline."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
76
+ gemini-1.5-pro,Gemini 1.5 Pro,Google,2024-05-14,,,moe,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemini 2.0 Flash Model Card (April 15 2025): non-thinking generation, default sampling. Same canonical setting applies to 1.5 / 2.0 family."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
77
+ gemini-2.0-flash,Gemini 2.0 Flash,Google,2025-02-01,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemini 2.0 Flash Model Card (April 15 2025): non-thinking generation, default sampling. Same canonical setting applies to 1.5 / 2.0 family."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
78
+ gemini-2.0-flash-lite,Gemini 2.0 Flash-Lite,Google,2025-02-25,,,moe,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemini 2.0 Flash Model Card (April 15 2025): non-thinking generation, default sampling. Same canonical setting applies to 1.5 / 2.0 family."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
79
+ gemini-2.0-pro,Gemini 2.0 Pro,Google,2025-02-05,,,moe,False,False,"{""context"": ""default (1M)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT (Google self-report). Gemini 2.0 Pro experimental."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
80
+ gemini-2.5-flash,Gemini 2.5 Flash,Google,2025-05-20,,,,True,False,"{""context"": ""default (1M)"", ""effort"": ""default"", ""harness"": ""AI Studio API (gemini-2.5-flash GA)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemini 2.5 Flash Model Card (Dec 2025) GA Thinking column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single attempt; multi-trial avg for smaller bench)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
81
+ gemini-2.5-pro,Gemini 2.5 Pro (GA),Google,2025-06-27,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""AI Studio API"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemini 2.5 Pro Model Card (GA, June 27 2025): pass@1 single attempt, AI Studio API model-id gemini-2.5-pro-preview-06-05 / gemini-2.5-pro GA, default sampling. Release date set to GA date 2025-06-27."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single attempt)"", ""temperature"": ""default"", ""tools"": ""none""}",True
82
+ gemini-3-flash,Gemini 3 Flash,Google,2025-11-18,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Google DeepMind /models/gemini/flash/ table: Gemini 3 Flash Thinking, no tools by default."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
83
+ gemini-3-pro,Gemini 3 Pro,Google,2025-11-18,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Google DeepMind /models/gemini/pro/ table: Gemini 3 Pro Thinking (High), no tools by default."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
84
+ gemini-3.1-pro,Gemini 3.1 Pro,Google,2026-02-19,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Google DeepMind /models/gemini/pro/ table: Gemini 3.1 Pro Thinking (High), no tools by default."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
85
+ gemini-diffusion,Gemini Diffusion,Google,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
86
+ gemma-2-27b,Gemma 2 27B,Google,2024-06-27,27000.0,27000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
87
+ gemma-2-2b,Gemma 2 2B,Google,2024-06-27,2000.0,2000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
88
+ gemma-2-9b,Gemma 2 9B,Google,2024-06-27,9000.0,9000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
89
+ gemma-3-12b,Gemma 3 12B,Google,2025-03-12,12000.0,12000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
90
+ gemma-3-1b,Gemma 3 1B,Google,2025-03-12,1000.0,1000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
91
+ gemma-3-1b-it,Gemma 3 1B,Google,2025,,,,,,"{""context"": ""32k"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Google Gemma 3 1B instruction-tuned."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
92
+ gemma-3-27b,Gemma 3 27B,Google,2025-03-12,27000.0,27000.0,Dense,False,True,"{""context"": ""default (128K)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report (arxiv:2503.19786) Table 6 IT column for 27B."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
93
+ gemma-3-4b,Gemma 3 4B,Google,2025-03-12,4000.0,4000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Gemma 3 tech report Table 6 IT instruction-tuned column."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
94
+ gemma-3n-e2b-instructed,Gemma 3n E2B Instructed,Google,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
95
+ gemma-3n-e2b-instructed-litert-preview,Gemma 3n E2B Instructed LiteRT (Preview),Google,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
96
+ gemma-3n-e4b-instructed,Gemma 3n E4B Instructed,Google,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
97
+ gemma-3n-e4b-instructed-litert-preview,Gemma 3n E4B Instructed LiteRT Preview,Google,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
98
+ gemma-4-26b-a4b,Gemma 4 26B A4B,Google,2026-04-22,26000.0,4000.0,MoE,True,True,"{""context"": ""default (128K E2B/E4B; 256K 31B/26B)"", ""effort"": ""default"", ""harness"": ""official (transformers/llama.cpp/MLX/SGLang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemma 4 HF blog benchmark table. Reasoning model with enable_thinking=True default."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
99
+ gemma-4-31b,Gemma 4 31B,Google,2026-04-22,31000.0,31000.0,Dense,True,True,"{""context"": ""default (128K E2B/E4B; 256K 31B/26B)"", ""effort"": ""default"", ""harness"": ""official (transformers/llama.cpp/MLX/SGLang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemma 4 HF blog benchmark table. Reasoning model with enable_thinking=True default."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
100
+ gemma-4-e2b,Gemma 4 E2B,Google,2026-04-22,2000.0,2000.0,Dense,True,True,"{""context"": ""default (128K E2B/E4B; 256K 31B/26B)"", ""effort"": ""default"", ""harness"": ""official (transformers/llama.cpp/MLX/SGLang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemma 4 HF blog benchmark table. Reasoning model with enable_thinking=True default."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
101
+ gemma-4-e4b,Gemma 4 E4B,Google,2026-04-22,4000.0,4000.0,Dense,True,True,"{""context"": ""default (128K E2B/E4B; 256K 31B/26B)"", ""effort"": ""default"", ""harness"": ""official (transformers/llama.cpp/MLX/SGLang)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Gemma 4 HF blog benchmark table. Reasoning model with enable_thinking=True default."", ""prompt_style"": ""default (enable_thinking=True)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
102
+ smollm2-1.7b,SmolLM2 1.7B Instruct,Hugging Face,2024-11-01,1700.0,1700.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
103
+ granite-3.3-8b-base,Granite 3.3 8B Base,IBM,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
104
+ granite-3.3-8b-instruct,Granite 3.3 8B,IBM,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
105
+ granite-4.0-1b,Granite-4.0-1B,IBM,2025,,,,,,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""IBM Granite 4.0 1B instruct."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
106
+ granite-4.0-h-1b,Granite-4.0-H-1B,IBM,2025,,,,,,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""IBM Granite 4.0 Hybrid 1B instruct."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
107
+ ibm-granite-4.0-tiny-preview,IBM Granite 4.0 Tiny Preview,IBM,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
108
+ exaone-4.0-32b,EXAONE 4.0 32B,LG AI Research,2025-07-15,32000.0,32000.0,"Hybrid (Local+Global Attention 3:1, QK-Reorder-LN)",True,True,"{""context"": ""128k"", ""effort"": ""default"", ""harness"": ""LG AI Research internal"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per LG AI Research arxiv:2507.11407 (Table 3): Reasoning mode is canonical (lead results). 32B max context 131,072 tokens."", ""prompt_style"": ""default"", ""sampling"": ""avg-of-N (n=32 AIME/HMMT, n=8 GPQA, n=4 LCB/Tau-bench, n=1 others)"", ""temperature"": ""0.6, top_p=0.95, presence_penalty=1.5 (Reasoning mode)"", ""tools"": ""none (per-bench override)""}",False
109
+ exaone-deep-32b,EXAONE-Deep-32B,LG AI Research,2025-03-18,32000.0,32000.0,dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Phi-4-reasoning paper Table 1 baseline (Microsoft self-test). Distinct from BP exaone-4.0-32b."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
110
+ lfm2.5-1.2b-instruct,LFM2.5-1.2B-Instruct,Liquid AI,2025-11,,,,,,"{""context"": ""32k"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Liquid AI LFM2.5 introduction blog. Instruct (non-thinking) variant."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
111
+ lfm2.5-1.2b-thinking,LFM2.5-1.2B-Thinking,Liquid AI,2026-01-06,1200.0,1200.0,Hybrid,True,True,,False
112
+ longcat-flash-chat,LongCat-Flash-Chat,Meituan,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
113
+ llama-3.1-405b-instruct,LLaMA-3.1 405B Instruct,Meta,2024-07-23,,,,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
114
+ llama-3.1-70b,Llama 3.1 70B Instruct,Meta,2024-07-23,70000.0,70000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
115
+ llama-3.1-8b,Llama 3.1 8B Instruct,Meta,2024-07-23,8000.0,8000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
116
+ llama-3.2-1b,Llama 3.2 1B Instruct,Meta,2024-09-25,1000.0,1000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
117
+ llama-3.2-1b-instruct,Llama 3.2 1B,Meta,2024-09,,,,,,"{""context"": ""128k"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Meta Llama 3.2 1B Instruct."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
118
+ llama-3.3-70b-instruct,Llama-3.3-70B-Instruct,Meta,2024-12-06,70000.0,70000.0,dense,False,True,"{""context"": ""default (128K)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Phi-4 paper Table 1 baseline."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
119
+ llama-4-behemoth,Llama 4 Behemoth,Meta,2025-04-05,2000000.0,288000.0,MoE,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official-meta"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""multimodal_input"": true, ""prompt_style"": ""zeroshot"", ""sampling"": ""pass@1"", ""temperature"": ""0.0"", ""tools"": ""none""}",False
120
+ llama-4-maverick,Llama 4 Maverick,Meta,2025-04-05,402000.0,17000.0,MoE,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official-meta"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""multimodal_input"": true, ""prompt_style"": ""zeroshot"", ""sampling"": ""pass@1"", ""temperature"": ""0.0"", ""tools"": ""none""}",True
121
+ llama-4-scout,Llama 4 Scout,Meta,2025-04-05,109000.0,17000.0,MoE,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official-meta"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""multimodal_input"": true, ""prompt_style"": ""zeroshot"", ""sampling"": ""pass@1"", ""temperature"": ""0.0"", ""tools"": ""none""}",False
122
+ muse-spark,Muse Spark,Meta,2026-04-01,,,,True,False,"{""effort"": ""default"", ""harness"": ""internal (Meta MSL)"", ""judge"": ""rule-based or gpt-oss-120b (per benchmark)"", ""mode"": ""thinking"", ""notes"": ""Per Meta MSL blog https://ai.meta.com/blog/introducing-muse-spark-msl/. Standard Thinking mode. Contemplating mode results flagged for review."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (single-pass)"", ""temperature"": ""default"", ""tools"": ""none (per-benchmark override where tools are required)""}",True
123
+ phi-3-14b,Phi-3 Medium 14B,Microsoft,2024-05-21,14000.0,14000.0,dense,False,True,"{""context"": ""128K"", ""effort"": ""n/a"", ""harness"": ""simple-evals"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Phi-4 paper Table 1 (Microsoft self-test baseline)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""0.5"", ""tools"": ""none""}",False
124
+ phi-3.5-mini-instruct,Phi-3.5-mini-instruct,Microsoft,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
125
+ phi-3.5-moe-instruct,Phi-3.5-MoE-instruct,Microsoft,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
126
+ phi-4,Phi-4,Microsoft,2025-01-01,14000.0,14000.0,Dense,False,True,"{""context"": ""default (16K)"", ""effort"": ""n/a"", ""harness"": ""OpenAI simple-evals"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Phi-4 paper (arxiv:2412.08905) Table 1, simple-evals framework, temperature 0.5."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""0.5"", ""tools"": ""none""}",False
127
+ phi-4-mini,Phi-4-mini,Microsoft,2025-03-01,3800.0,3800.0,Dense,False,True,"{""context"": ""default (128K)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Phi-4-mini-instruct HF model card."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
128
+ phi-4-reasoning,Phi-4-reasoning,Microsoft,2025-04-30,14000.0,14000.0,Dense,True,True,"{""context"": ""default (32K)"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Phi-4-reasoning paper (arxiv:2504.21318) Tables 1-2. Thinking model, temp 0.8, max_seq_len 31k."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg of 5)"", ""temperature"": ""0.8"", ""tools"": ""none""}",False
129
+ phi-4-reasoning-plus,Phi-4-reasoning-plus,Microsoft,2025-04-30,14000.0,14000.0,Dense,True,True,"{""context"": ""default (32K)"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Phi-4-reasoning paper. Same as phi-4-reasoning + GRPO RL on top."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg of 5)"", ""temperature"": ""0.8"", ""tools"": ""none""}",False
130
+ minimax-m2,MiniMax-M2,MiniMax,2025-10-01,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per MiniMax M2 model card."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
131
+ minimax-m2.1,MiniMax M2.1,MiniMax,2025-12-01,,,,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per MiniMax M2.5 model card (referenced as predecessor)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
132
+ minimax-m2.5,MiniMax-M2.5,MiniMax,2026-04-22,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per MiniMax M2.5 model card."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
133
+ minimax-m2.7,MiniMax M2.7,MiniMax,2026-04-01,,,,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per GLM-5.1 blog (cross-model)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
134
+ codestral-25.01,Codestral 25.01,Mistral,2025-01-15,22000.0,22000.0,Dense,False,True,,False
135
+ devstral-2,Devstral 2,Mistral,2025-12-01,123000.0,,Dense,False,True,,False
136
+ ministral-8b,Ministral 8B Instruct 2410,Mistral,2024-10-16,8000.0,8000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
137
+ mistral-large-2,Mistral Large 2,Mistral,2024-07-24,123000.0,123000.0,dense,False,True,"{""context"": ""128K"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Command A paper (third-party Cohere reproduction baseline)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
138
+ mistral-large-3,Mistral Large 3,Mistral,2025-12-01,675000.0,,MoE,False,True,,False
139
+ mistral-medium-3,Mistral Medium 3,Mistral,2025-05-01,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""mistral-eval"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""multimodal_input"": true, ""notes"": ""Per Mistral Medium 3 blog: https://mistral.ai/news/mistral-medium-3. \""Same internal evaluation pipeline\"" across all benchmarks."", ""prompt_style"": ""mixed (5-shot CoT for GPQA Diamond / MMLU Pro per Mistral blog; 0-shot for others)"", ""sampling"": ""pass@1"", ""temperature"": ""0.0"", ""tools"": ""none""}",False
140
+ mistral-small-24b,Mistral Small 24B Instruct,Mistral,2025-01-30,24000.0,24000.0,dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
141
+ mistral-small-3.1,Mistral Small 3.1,Mistral,2025-03-01,24000.0,24000.0,Dense,False,True,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""mistral-eval"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""multimodal_input"": true, ""prompt_style"": ""mixed (5-shot CoT for MMLU/Pro/GPQA per HF MC; default for others)"", ""sampling"": ""pass@1"", ""temperature"": ""0.0"", ""tools"": ""none""}",False
142
+ codestral-22b,Codestral-22B,Mistral AI,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
143
+ mistral-small-3-24b-instruct,Mistral Small 3 24B Instruct,Mistral AI,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
144
+ pixtral-12b,Pixtral-12B,Mistral AI,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
145
+ kimi-k2,Kimi K2,Moonshot AI,2025-07-16,,,MoE,False,True,"{""context"": ""default (256K)"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per Kimi K2 tech report (arxiv:2507.20534) Table 3 Kimi-K2-Instruct column. Non-thinking model."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg of 4-64 trials per benchmark)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
146
+ kimi-k2-thinking,Kimi K2 Thinking,Moonshot AI,2025-11-01,,,MoE,True,True,"{""context"": ""default (256K)"", ""effort"": ""default (96k thinking-token budget for HLE/AIME/HMMT/GPQA; 128k for IMO/LCB/OJ)"", ""harness"": ""official"", ""judge"": ""rule-based (HLE uses o3-mini)"", ""mode"": ""thinking"", ""notes"": ""Per Kimi K2 Thinking blog (moonshotai.github.io/Kimi-K2/thinking.html). Thinking model."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg@32 for AIME/HMMT no-tools)"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
147
+ kimi-k2.5,Kimi K2.5,Moonshot AI,2026-01-01,,,MoE,True,True,"{""context"": ""256K"", ""effort"": ""default (96k for reasoning, 64k for vision)"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Kimi K2.5 model card (HF moonshotai/Kimi-K2.5). Thinking mode, t=1.0 top_p=0.95."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg@32 AIME/HMMT, avg@8 GPQA, avg@3 vision, avg@4 Seal/WideSearch)"", ""temperature"": ""1.0"", ""tools"": ""none (per-bench override)""}",True
148
+ kimi-k2.6,Kimi K2.6,Moonshot AI,2026-04-22,,,MoE,True,True,"{""context"": ""262144 (256K)"", ""effort"": ""max (98304 generation tokens)"", ""harness"": ""official (in-house SWE-agent-derived for SWE-Bench)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Kimi K2.6 model card (HF moonshotai/Kimi-K2.6). Thinking mode, max effort, t=1.0 top_p=1.0."", ""prompt_style"": ""default"", ""sampling"": ""pass@1 (avg@10 coding, avg@3 vision)"", ""temperature"": ""1.0"", ""tools"": ""none (per-bench override)""}",True
149
+ nemotron-ultra-253b,Nemotron Ultra 253B,NVIDIA,2025-04-10,253000.0,,MoE,True,True,"{""context"": ""32k"", ""effort"": ""default"", ""harness"": ""NVIDIA eval pipeline (transformers/vLLM)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per NVIDIA HF model card https://huggingface.co/nvidia/Llama-3_1-Nemotron-Ultra-253B-v1: Reasoning On mode, temperature=0.6, top_p=0.95, 32k context, benchmarks run up to 16x and averaged."", ""prompt_style"": ""custom prompt templates per benchmark"", ""sampling"": ""pass@1 (avg of up to 16 benchmark runs)"", ""temperature"": ""0.6 (thinking mode)"", ""tools"": ""none (benchmark-specified)""}",False
150
+ openthinker2-32b,OpenThinker2-32B,Open Thoughts,2025-04-04,32000.0,32000.0,dense,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Phi-4-reasoning paper Table 1 baseline (Microsoft self-test)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
151
+ gpt-3.5-turbo,GPT-3.5 Turbo (0125),OpenAI,2024-01-25,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""OLMES (Allen AI)"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OLMo 2 paper Table 7 Instruct (instruction-tuned variant)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
152
+ gpt-4-turbo,GPT-4 Turbo,OpenAI,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
153
+ gpt-4.1,GPT-4.1,OpenAI,2025-04-14,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. non-reasoning model."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
154
+ gpt-4.1-mini,GPT-4.1 mini,OpenAI,2025-04-14,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. non-reasoning model."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
155
+ gpt-4.1-nano,GPT-4.1 nano,OpenAI,2025-04-14,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. non-reasoning model."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
156
+ gpt-4.5,GPT-4.5,OpenAI,2025-02-27,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-4.1 blog table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
157
+ gpt-4o,GPT-4o (2024-11-20),OpenAI,2024-11-20,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-4.1 blog table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
158
+ gpt-4o-0513,GPT-4o (2024-05-13),OpenAI,2024-05-13,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per DeepSeek V3 paper Table 6."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",True
159
+ gpt-4o-mini,GPT-4o mini,OpenAI,2024-07-18,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per OpenAI GPT-4.1 blog table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
160
+ gpt-5,GPT-5,OpenAI,2025-08-01,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. high reasoning effort."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
161
+ gpt-5-mini,GPT-5 mini,OpenAI,2025-07-10,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5 developer blog https://openai.com/index/introducing-gpt-5-for-developers/"", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
162
+ gpt-5-nano,GPT-5 nano,OpenAI,2025-07-10,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5 developer blog https://openai.com/index/introducing-gpt-5-for-developers/"", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
163
+ gpt-5.1,GPT-5.1,OpenAI,2025-11-13,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5.1 dev blog: \""high\"" reasoning effort (not xhigh like 5.2/5.4/5.5)."", ""prompt_style"": ""default (with prompt adjustment for τ²-bench)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
164
+ gpt-5.2,GPT-5.2,OpenAI,2025-12-11,,,,True,False,"{""context"": ""default"", ""effort"": ""xhigh"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-5.2 blog: reasoning effort=xhigh, research environment."", ""prompt_style"": ""default (with prompt adjustment for τ²-bench)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
165
+ gpt-5.2-codex,GPT-5.2-Codex,OpenAI,2025-07-10,,,,True,False,"{""context"": ""default"", ""effort"": ""xhigh"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5.3-Codex blog https://openai.com/index/introducing-gpt-5-3-codex/ — shows GPT-5.2-Codex (xhigh) as cross-model baseline"", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
166
+ gpt-5.3-codex,GPT-5.3-Codex,OpenAI,2026-02-05,,,,True,False,"{""context"": ""default"", ""effort"": ""xhigh"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-5.3-codex blog: reasoning effort=xhigh, research environment."", ""prompt_style"": ""default (with prompt adjustment for τ²-bench)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",False
167
+ gpt-5.4,GPT-5.4,OpenAI,2026-03-01,,,,True,False,"{""context"": ""default"", ""effort"": ""xhigh"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-5.4 blog: reasoning effort=xhigh, research environment (may differ from production ChatGPT)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
168
+ gpt-5.5,GPT-5.5,OpenAI,2026-04-22,,,,True,False,"{""context"": ""default"", ""effort"": ""xhigh"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-5.5 blog: reasoning effort=xhigh, research environment (may differ from production ChatGPT)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
169
+ gpt-oss-120b,gpt-oss-120B,OpenAI,2025-08-05,116800.0,5100.0,MoE,True,True,"{""context"": ""default (128K)"", ""effort"": ""high"", ""harness"": ""official (harmony chat format)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-oss model card (arxiv:2508.10925) Table 3, high reasoning effort. Default with-tools for agentic; AIME/GPQA also has no-tools variant in same table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""with tools (per-bench override; harmony chat format default)""}",True
170
+ gpt-oss-20b,gpt-oss-20B,OpenAI,2025-08-05,20900.0,3600.0,MoE,True,True,"{""context"": ""default (128K)"", ""effort"": ""high"", ""harness"": ""official (harmony chat format)"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI gpt-oss model card (arxiv:2508.10925) Table 3, high reasoning effort. Default with-tools for agentic; AIME/GPQA also has no-tools variant in same table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""with tools (per-bench override; harmony chat format default)""}",False
171
+ o1-high,OpenAI o1 (high),OpenAI,2024-12-17,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-4.1 blog table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
172
+ o1-mini,o1-mini,OpenAI,2024-09-12,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per Phi-4-reasoning paper Table 1 baseline (Microsoft self-test)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
173
+ o3-high,o3 (high),OpenAI,2025-04-16,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. high reasoning effort."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
174
+ o3-mini-high,o3-mini (high),OpenAI,2025-01-31,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-4.1 blog table."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
175
+ o4-mini-high,o4-mini (high),OpenAI,2025-04-16,,,,True,False,"{""context"": ""default"", ""effort"": ""high"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per OpenAI GPT-5 dev blog. high reasoning effort."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
176
+ minicpm-sala,MiniCPM-SALA,OpenBMB,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
177
+ sarvam-30b,Sarvam-30B,Sarvam AI,,,,,False,True,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
178
+ internlm3-8b,InternLM3-8B-Instruct,Shanghai AI Lab,2025-01-15,8000.0,8000.0,Dense,False,True,"{""effort"": ""default"", ""harness"": ""OpenCompass"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per https://github.com/InternLM/InternLM README: evaluated using OpenCompass. Values marked * = Thinking Mode (excluded from non-thinking canonical). Most benchmarks 0-shot."", ""prompt_style"": ""0-shot (default per benchmark)"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none""}",False
179
+ falcon3-10b,Falcon3-10B-Instruct,TII,2025-01-01,10000.0,10000.0,Dense,False,True,,False
180
+ glm-4.6,GLM-4.6,Zhipu AI,2025-09-01,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per GLM-4.7 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
181
+ glm-4.7,GLM-4.7,Zhipu AI,2025-12-01,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per GLM-5 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
182
+ glm-5,GLM-5,Zhipu AI,2026-03-15,,,,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per GLM-5 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
183
+ glm-5.1,GLM-5.1,Zhipu AI,2026-04-07,,,,True,True,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per GLM-5.1 blog."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
184
+ grok-1.5,Grok-1.5,xAI,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
185
+ grok-2,Grok-2,xAI,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
186
+ grok-2-mini,Grok-2 mini,xAI,,,,,False,False,"{""context"": ""unknown"", ""effort"": ""n/a"", ""harness"": ""unknown"", ""judge"": ""unknown"", ""mode"": ""unknown"", ""notes"": ""Model added from third-party aggregator (llm-stats.com); canonical setting not established."", ""prompt_style"": ""unknown"", ""sampling"": ""unknown"", ""temperature"": ""unknown"", ""tools"": ""unknown""}",False
187
+ grok-3-beta,Grok 3 Beta,xAI,2025-02-19,,,,False,False,"{""context"": ""default"", ""effort"": ""n/a"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""non-thinking"", ""notes"": ""Per xAI Grok 3 blog: canonical = non-reasoning mode (the cross-model table without 'Think' label). Reasoning ON values not used."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
188
+ grok-4,Grok 4,xAI,2025-07-09,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per xAI Grok 4 blog: standard chart values (no Heavy, no with-Python tool unless noted)."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
189
+ grok-4.1,Grok 4.1,xAI,2025-11-17,,,,True,False,"{""context"": ""default"", ""effort"": ""default"", ""harness"": ""official"", ""judge"": ""rule-based"", ""mode"": ""thinking"", ""notes"": ""Per xAI Grok 4.1 blog: standard chart values. Hallucination/FActScore use with-web-search but those are not stored in BP."", ""prompt_style"": ""default"", ""sampling"": ""pass@1"", ""temperature"": ""default"", ""tools"": ""none (per-bench override)""}",True
190
+ grok-4.20,Grok 4.20,xAI,2026-03-09,,,,True,False,,False
data/models.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0cde346c85c4465891d465647c1739bf4f3fa0dd1a2a3e69c6eacc2dfd2467e6
3
+ size 21161
data/score_matrix_paper_wide.csv ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ model_id,aa_intelligence_index,aa_lcr,aider_polyglot_diff,aider_polyglot_whole,aime_2024,aime_2025,aime_2026,alpacaeval_2,arc_agi_1,arc_agi_2,arena_hard,artifactsbench,babyvision,beyond_aime,bfcl,bfcl_v3,bigbench_hard,bigcodebench,bird_sql,browsecomp,browsecomp_long_context_128k,browsecomp_zh,brumo_2025,bullshit_pushback,c_eval,charxiv_descriptive,charxiv_reasoning,chatbot_arena_elo,chinese_simpleqa,cluewsc,cmimc_2025,cnmo_2024,codeforces_rating,collie,complexfuncbench,critpt,cybergym,deepsearchqa_acc,drop,erqa,facts_grounding,factscore,finance_agent,finsearchcomp_global,frames,frontiermath,frontiermath_tier4,gaia,gdpval_aa_elo,global_piqa,gpqa_diamond,gpqa_main,graphwalks_bfs_0k_128k,graphwalks_parents_0k_128k,gsm8k,healthbench,hle,hle_text,hle_tools,hmmt_feb_2025,hmmt_feb_2026,hmmt_nov_2025,humaneval,ifbench,ifeval,imo_answerbench,infobench,internal_api_if_hard,livebench,livecodebench,longbench_v2,longfact_concepts,longfact_objects,math,math_500,matharena_apex_2025,mathvision,mathvista,mbpp_plus,mcpatlas,mcpmark,mgsm,mmlu_pro,mmmlu,mmmu,mmmu_pro,mrcr_v1,mrcr_v2,mrcr_v2_2needle_128k,mrcr_v2_8needle,mt_aime_2024,multi_if,multi_swe_bench,multichallenge,multichallenge_o3mini_grader,multipl_e_avg,nl2repo_bench,ojbench,omnidocbench,omnidocbench_1.5,osworld,popqa,repoqa,safety,scicode,screenspot_pro,simpleqa,simpleqa_verified,smt_2025,supergpqa,swe_bench_multilingual,swe_bench_pro,swe_bench_verified,swelancer,swelancer_freelance_dollars,tau2_bench_airline,tau2_bench_retail,tau2_bench_telecom,tau3_bench,tau_bench_airline,tau_bench_retail,terminal_bench,terminal_bench_1,terminal_bench_hard,toolathlon,truthfulqa,usamo_2025,vending_bench_2,vibe_eval,video_mme,video_mmmu,widesearch,xbench_deepsearch
2
+ claude-3.5-sonnet,,,45.3,,16.0,,,,,,,,,,,,93.1,,,,,,,,76.7,,,,51.3,85.4,,13.1,717.0,,,,,,88.3,,,,,,72.5,,,,,,65.0,65.0,,,96.4,,,,,,,,81.7,,86.5,,93.9,,,36.3,41.0,,,78.3,78.3,,,67.7,,,,91.6,78.0,82.1,70.4,,,,,,,,,,,,,,,,,,,,,,28.4,,,,,,50.8,,,,,,,48.8,71.5,,,,,,,,,,,,
3
+ claude-3.7-sonnet,,,,,23.3,54.8,,,28.6,0.7,63.9,,,,,,,,,,,,65.83,54.72,,,,1372.0,,,,,,,,,,,,,,,,,,3.1,,,,,78.2,84.8,,,,,,,,31.7,,,,,90.8,,,,,52.6,,,,,82.2,,,,,,,,,85.9,75.0,,,,,,80.0,,,,,,,,,,,,,,,,,,56.6,,,,62.3,,,,,,,58.4,81.2,35.2,35.2,,,,3.65,,,,,,
4
+ claude-haiku-4.5,,,,,,96.3,,,47.7,4.0,,,,,,,,,,,,,,87.27,,,,,,,,,,,,,,,,,,,,,,,,,,,73.0,73.0,,,,,,,,,,,,,,,,,61.0,,,,,,,,,,,,,,,83.0,73.2,,,,,,,,,,,,,,,,50.7,,,,,,,,,,,39.45,73.3,,,63.6,83.2,83.0,,,,41.0,,,,,,,,,,,
5
+ claude-opus-4,,,,,,75.5,,,35.7,8.6,59.7,,,,,,,,,,,,,32.73,,,,,,,,,,,,0.3,,,,,,,,,,4.5,,,,,79.6,79.6,,,,,7.1,,,15.9,,,,,89.7,,,,,47.4,,,,,94.4,,,,,,,,86.6,88.8,76.5,,,,,,,,,,,,,,,,,,,,,,22.8,,,,,,72.5,,,,,,,59.6,81.4,43.2,39.0,,,,,,,,,,
6
+ claude-opus-4.1,,,,,,78.0,,,,,,,,,,,,,,,,,,40.0,,,,,,,,,,,,,,,,,,,,,,5.9,,,,,80.9,80.9,,,,,,,,,,,,,,,,,61.4,,,,,,,,,,,,,,,89.5,77.1,,,,,,,,,,,,,,,,44.4,,,,,,,,,,,22.7,74.5,,,,,,,56.0,82.4,43.3,43.8,,,,,,,,,,
7
+ claude-opus-4.5,,71.3,92.4,,,92.8,93.3,,84.0,37.6,,68.5,16.2,69.0,,,,,,67.8,,62.4,,90.91,,92.7,67.2,1469.0,,,,,1701.0,79.8,,,50.6,76.1,,48.3,,91.1,58.8,,84.7,20.7,,,1416.0,91.6,87.0,87.0,92.0,96.2,,36.3,30.8,43.2,43.2,92.9,,91.7,,58.0,90.9,78.5,,,76.0,84.8,64.4,98.8,99.0,,,1.6,77.1,80.6,,62.3,42.3,,89.3,90.8,80.7,70.6,,,,56.2,,,52.8,54.2,,,43.2,54.6,,0.123,66.3,,,,49.5,45.7,,48.6,,70.6,77.5,55.4,80.9,56.1,,,88.9,98.2,,,88.9,59.3,,,43.5,,,4967.06,70.3,77.6,84.4,76.2,
8
+ claude-opus-4.6,53.0,71.0,,,,95.6,95.6,,93.0,68.8,,,14.8,,,,,,,83.7,,,,92.73,,,61.5,1496.0,76.4,,,,,,,12.6,73.8,80.6,,,,,60.1,,,40.7,,,1606.0,,91.3,91.3,,,,,40.0,53.1,,,96.2,96.3,,53.0,,75.3,,,,88.8,,,,,,34.5,71.2,,,75.8,56.7,,89.1,91.1,,73.9,92.9,84.0,,93.0,,,,,,,49.8,60.3,,,72.7,,,,52.0,,,46.2,,,77.8,53.4,80.8,,,,91.9,99.3,72.4,,,65.4,,,47.2,,42.3,8017.59,,,,,
9
+ claude-opus-4.7,,,,,,,,,92.0,75.8,,,,,,,,,,79.3,,,,80.0,,,82.1,,,,,,,,,,73.1,,,,,,64.4,,,43.8,22.9,,,,94.2,94.2,,,,,46.9,54.7,,,,,,,,,,,,,,,,,,,,,,77.3,,,,91.5,,,,,,,,,,,,,,,,,78.0,,,,,,,,,,,64.3,87.6,,,,,,,,,69.4,,,,,,,,,,,
10
+ claude-sonnet-4,57.0,65.0,,,43.4,70.5,,,40.0,5.9,51.6,57.3,,,,,,,,12.2,,29.1,,25.45,,,,,,,,,,,,,,,,,,,,42.0,,,,68.3,,,75.4,75.4,,,,,5.8,,,15.9,,,,55.0,87.6,,,,,66.0,,,,,94.0,,,,,,,,84.0,86.5,74.4,,,,,,,,35.7,,,,,,,,,,,,40.0,,15.9,,,,56.9,42.7,72.7,,,63.0,83.8,49.6,,60.0,80.5,35.5,36.4,30.0,,,,,,,,,64.6
11
+ claude-sonnet-4.5,63.0,66.0,82.2,,,87.0,82.5,,70.9,13.6,,61.5,,57.0,,,,,,43.9,,42.4,90.83,65.45,,,68.5,,,,66.88,,1480.0,77.3,,,,36.3,,,,90.6,54.5,60.8,78.7,,,71.2,1276.0,90.1,83.4,83.4,80.5,99.0,,44.2,17.7,32.0,,79.2,,81.7,,57.0,,65.8,,,67.9,64.0,62.0,98.5,98.8,,,1.56,,,,43.8,32.1,,88.2,89.1,77.8,63.4,,47.1,,47.1,,,44.3,57.3,,,39.9,30.4,0.145,0.145,61.4,,,,44.7,36.2,,29.3,83.96,65.5,68.0,48.4,77.2,45.7,,70.0,86.2,98.0,,70.0,86.2,50.0,51.0,33.3,38.9,,,3839.0,,,77.8,65.1,66.0
12
+ claude-sonnet-4.6,,,,,,95.6,,,86.5,58.3,,,,,,,,,,74.7,,,,94.55,,,,,,,,,,,,,65.2,,,,,,63.3,,,32.4,,,1633.0,,89.9,89.9,,,,,33.2,49.0,,,,,,,,,,,,,,,,,,,,,,61.3,,,,89.3,,74.5,,84.9,,,,,,,,,,,,,72.5,,,,47.0,,,,,,,,79.6,,,,91.7,97.9,,,,59.1,,,,,,7204.14,,,,,
13
+ command-a,,,,,,,,,,,,,,,63.8,,,45.4,59.5,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,50.8,,,,,,,,,,,82.9,,89.7,,94.9,,,26.9,,,,80.0,82.0,,,,86.2,,,,68.9,,,,,,,,,,,,,73.1,,,,,,,92.6,,,,,,,,,,26.8,,,,,,,,,,,,,,,,,,,,
14
+ deepseek-r1,,,53.3,71.6,79.8,70.0,,87.6,18.3,1.3,92.3,,,42.4,,,,,,8.9,,35.7,80.83,12.73,91.8,,,1398.0,63.7,92.8,,78.8,2029.0,34.2,,1.1,,,92.2,,,,,,82.5,,,,,,71.5,81.0,,,,,8.5,17.7,,41.7,,,,38.0,83.3,,,,,65.9,,,,,97.3,,,,,,,,84.0,,,,,,,,91.4,,,,,,,,,,,,,,,,30.1,,,60.5,30.5,,49.2,,,,,,,,,,5.7,,,,4.76,,,,,,
15
+ deepseek-r1-0528,,,71.6,,91.4,87.5,,,21.2,1.1,,,,,,,,,,,,,92.5,,,,,1422.0,,,69.38,86.9,1930.0,,,,,,,,,,,,83.0,,,,,,81.0,,,,,,17.7,,,79.4,,,,,79.1,,,,,73.3,,,,,,1.04,,,,,,,85.0,,,,,,,,,,,,,,,,,,,,,,,,27.8,,83.02,,,,57.6,,,,,,,53.5,63.9,,,,,,30.06,,,,,,
16
+ deepseek-r1-distill-llama-70b,,,,,70.0,56.3,,,,,60.6,,,,,49.3,,,,,,,,,,,,,,,,,1633.0,,,,,,,,,,,,,,,,,,65.2,65.2,,,,,,,,33.3,,,,,79.3,,,,54.5,54.5,,,,,94.5,,,,,,,,,69.6,,,,,,,29.3,57.6,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
17
+ deepseek-v2-0506,,,,,4.6,,,,,,,,,,,,,,,,,,,,78.6,,,,48.5,89.9,,2.8,,,,,,,83.0,,,,,,66.9,,,,,,35.3,,,,,,,,,,,,69.3,,57.7,,,,,18.8,31.6,,,,56.3,,,,,,,,58.5,,,,,,,,,,,,,,,,,,,,,,,,9.0,,,,,,,,,,,,,,,,,,,,,,,,,,
18
+ deepseek-v2.5-0905,,,18.2,,16.7,,,,,,76.2,,,,,,,,,,,,,,79.5,,,,54.1,90.4,,10.8,,,,,,,87.8,,,,,,65.4,,,,,,41.3,,,,95.1,,,,,,,,77.4,,80.6,,,,,29.2,35.4,,,74.7,74.7,,,,,,,,66.2,,,,,,,,,,,,,,,,,,,,,,,,10.2,,,,,,22.6,,,,,,,,,,,,,,,,,,,,
19
+ deepseek-v3,,,49.6,49.6,39.2,,,,,,85.5,,,,72.1,,,50.0,50.5,,,,,,86.5,,,1358.0,64.8,90.9,,43.2,1134.0,,,,,,91.6,,,,,,73.3,,,,,,59.1,53.6,,,89.3,,,,,,,,82.6,,86.1,,94.3,,,40.5,48.7,,,68.5,90.2,,,,86.5,,,,75.9,,,,,,,,59.4,,,,,,,,,,,,91.2,,,,24.9,,,,,,42.0,,,,,,,,,,,,,,,,,,,,
20
+ deepseek-v3.2,57.0,64.3,74.5,74.5,,93.1,95.1,,57.0,4.0,,55.8,,,,,,,,51.4,,65.0,96.67,14.55,,,,,,,83.75,,2386.0,,,,17.3,60.9,,,,,,26.2,,22.1,,63.5,,,82.4,82.4,,,,46.9,25.1,40.8,40.8,92.5,79.9,90.2,,54.0,,78.3,,,63.1,83.3,59.8,,,,,2.08,,,,62.2,38.0,,85.0,,,,,,,,,,30.6,,,,,54.7,,,,,,,38.9,,,,87.74,,70.2,,73.1,,,63.8,81.1,96.2,69.2,,,46.4,,35.4,35.2,,,1034.0,,,,32.5,71.0
21
+ deepseek-v4-flash,,,,,,,,,,,,,,,,,,,,73.2,,,,,,,,,78.9,,,,3052.0,,,,,,,,,,,,,,,,1395.0,,88.1,88.1,,,,,34.8,45.1,,,94.8,,,,,88.4,,,,91.6,,,,,,33.0,,,,69.0,,,86.2,,,,78.7,,,,,,,,,,,,,,,,,,,,34.1,34.1,,,73.3,52.6,79.0,,,,,,,,,56.9,,,47.8,,,,,,,,
22
+ deepseek-v4-pro,,,,,,,,,,,,,,,,,,,,83.4,,,,,,,,,84.4,,,,3206.0,,,,,,,,,,,,,,,,1554.0,,90.1,90.1,,,,,37.7,48.2,,,95.2,,,,,89.8,,,,93.5,,,,,,38.3,,,,73.6,,,87.5,,,,83.5,,,,,,,,,,,,,,,,,,,,57.9,57.9,,,76.2,55.4,80.6,,,,,,,,,67.9,,,51.8,,,,,,,,
23
+ doubao-seed-2.0-pro,,,80.0,,,98.3,94.2,,85.4,37.5,,66.6,60.6,86.5,,,,,,77.3,,82.4,,,,93.5,80.5,,,,,,3020.0,93.9,,,,77.4,,68.5,,71.2,,,84.5,,,,,92.3,88.9,,68.9,97.6,,57.7,32.4,54.2,,97.3,,93.3,,,,89.3,,,,87.8,63.8,92.8,92.9,,,20.3,88.8,89.8,,,54.7,,87.0,88.1,85.4,78.2,,,,54.0,,,45.2,68.3,,,27.9,,,0.099,,,,,48.5,,,36.0,,68.7,71.7,46.9,76.5,49.4,,,90.4,94.2,,,,55.8,,,,,,,81.4,89.5,86.9,74.7,
24
+ gemini-1.5-flash,,,,,,,,,,,,,,,,,85.5,,45.6,,,,,,,,,,,,,,,,,,,,,,82.9,,,,,,,,,,51.0,51.0,,,86.2,,,,,,,,74.3,,,,,,,30.7,,,,77.9,,,,65.8,,,,82.6,67.3,,62.3,,71.9,,,,,,,,,,,,,,,,,,,,8.6,,,,,,,,,,,,,,,,,,,,,,48.9,76.1,,,
25
+ gemini-1.5-pro,,,,,,,,,,0.8,,,,,,,89.2,,54.4,,,,,,,,,,,,,,,,,,,,74.9,,80.0,,,,,,,,,,59.1,59.1,,,90.8,,,,,,,,84.1,,,,,,,34.2,,,,86.5,,,,68.1,,,,87.5,75.8,,65.9,,82.6,,,,,,,,,,,,,,,,,,,,24.9,,,,,,,,,,,,,,,,,,,,,,53.9,78.6,,,
26
+ gemini-2.0-flash,,,,,,,,,,1.3,50.0,,,,,,,45.9,58.7,,,,,,,,,,,,,,,,,,,,,,84.6,,,,,,,,,,60.1,74.2,,,,,,,,,,,,,,,,,,34.5,,,,90.9,,,,,,,,,77.6,,71.7,,70.5,,,,73.3,,,,,,,,,,,,,,,,29.9,,,,,,,,,,,,,,,,,,,,,,56.3,,,,
27
+ gemini-2.5-flash,,,61.9,61.9,,72.0,,,33.3,2.5,83.9,,,,,,,,,,,,83.33,14.55,,,63.7,,,,51.88,,1995.0,,,1.1,,,,,85.3,,,,,,,,,90.2,82.8,82.8,,,,,11.0,11.0,,64.2,,,,,84.3,,,,46.9,63.9,,,,,,,,,,3.4,,,81.1,86.6,79.7,66.7,74.0,54.3,,,88.0,,,,,,,,0.154,0.154,,,,,,3.9,26.9,28.1,75.47,,,,60.4,,,,,,,,,16.9,16.8,,3.7,,,549.0,65.4,,79.2,,
28
+ gemini-2.5-pro,60.0,66.0,82.2,82.2,92.0,88.0,,,27.6,4.9,90.8,57.7,,58.8,,,,,,9.9,,32.2,90.0,23.64,,,69.6,1448.0,,,58.13,,2001.0,62.5,,2.0,,,,,87.8,,,42.6,,14.1,,60.2,,91.5,86.4,86.4,,,,,21.6,21.6,,82.5,,80.0,,49.0,91.5,,,,57.5,69.0,,,,,,,,,,8.8,,,86.0,89.5,82.0,68.0,93.0,58.0,,16.4,92.0,,,,,,,,0.145,0.145,,,,,43.0,11.4,54.0,54.5,84.91,65.3,,,59.6,,,,,54.0,,,,32.6,25.3,25.0,10.5,,24.0,574.0,67.2,86.9,83.6,,56.0
29
+ gemini-3-flash,,,,,,95.2,,,84.7,33.6,,,,,,,,,,,,,100.0,,,,80.3,1473.0,,,90.62,,,,,,,,,,61.9,,,,,35.6,,,,92.8,90.4,90.4,,,,,33.7,43.5,,,,93.33,,,,,,,73.0,90.8,,,,,,15.62,,,,57.4,,,,91.8,,81.2,,67.2,,22.1,,,,,,,,,0.121,0.121,,,,,,69.1,,68.7,92.92,,,34.63,78.0,,,,,,,,,47.6,,,49.4,,,3635.0,,,86.9,,
30
+ gemini-3-pro,,65.3,94.2,,,95.0,90.6,,85.0,31.1,,58.4,49.7,83.0,,,,,,59.2,,66.8,98.33,,,94.4,81.4,,,,90.0,,2708.0,95.0,,9.1,39.9,63.2,,70.5,70.5,92.6,,,81.9,,,,1195.0,93.4,91.9,91.9,79.9,99.7,,37.9,37.5,46.9,45.8,97.5,,93.3,,70.4,93.5,83.3,,,73.5,90.7,68.2,98.5,98.1,,,24.5,86.1,89.8,,54.1,53.9,,90.1,91.8,87.2,81.0,,77.0,,79.7,,,50.2,64.2,,,34.2,68.5,0.115,0.115,,,,,56.0,72.7,,72.1,93.4,74.0,65.0,43.3,76.2,44.3,,,85.3,98.0,,,,56.9,,39.0,36.4,,,5478.0,77.7,88.4,87.6,57.0,
31
+ gemini-3.1-pro,57.0,,,,,,98.2,,98.0,77.1,,,51.6,,,,,,,85.9,,,,,,,80.2,1493.0,85.9,,,,3052.0,,,17.7,38.8,60.2,,,,,59.7,,,36.9,16.7,,1314.0,,94.3,94.3,,,,,44.4,51.4,,,94.7,94.8,,,,81.0,,,,91.7,,,,,,60.9,89.8,,,69.2,55.9,,91.0,92.6,,80.5,76.3,84.9,,26.3,,,,,,,33.4,70.7,,,,,,,59.0,,,75.6,,,76.9,54.2,80.6,,,,90.8,99.3,67.1,,,68.5,,,48.8,,,911.21,,,,,
32
+ gemma-2-27b,,,,,,,,49.0,,,,,,,,,72.7,,46.7,,,,,,,,,,,,,,,,,,,,67.5,,62.4,,,,,,,,,,34.3,,,,80.7,,,,,,,,51.8,,63.2,,,,,20.4,,,,55.6,,,,,,,,,56.9,,,,,,,,,,,,,,,,,,,33.9,,75.9,,,9.2,,,,,,,,,,,,,,,,,,,64.6,,,,,,,
33
+ gemma-2-9b,,,,,,,,43.7,,,,,,,,,64.9,,33.8,,,,,,,,,,,,,,,,,,,,58.8,,62.0,,,,,,,,,,28.8,31.3,,,79.7,,,,,,,,40.2,,69.9,,87.2,,,10.8,,,,49.4,,,,,,,,,46.8,,,,,,,,,,,,,,,,,,,28.3,,75.5,,,5.3,,,,,,,,,,,,,,,,,,,61.4,,,,,,,
34
+ gemma-3-1b-it,,,,,,1.0,,20.4,,,,,,,,16.64,39.1,,6.4,,,,,,,,,,,,,,,,,,,,25.1,,36.4,,,,,,,,,,24.24,19.2,,,42.15,,,,,,,,41.5,20.47,63.25,,,,,1.9,,,,48.0,45.2,,,,,,,,14.04,,,,,,,,,44.31,,,,,,,,,,9.6,,70.2,,,2.2,,,,,,,,,,,,,,,,,,,43.8,,,,,,,
35
+ gemma-3-27b,,,,,,,,63.4,,,69.9,,,,,,83.7,,54.4,,,,,,,,,1338.0,,,,,,,,,,,69.2,,74.9,,,,,,,,,,42.4,42.4,,,,,,,,,,,,,,,,,,29.7,,,,89.0,,,,,,,,,67.5,,64.9,,,,,,,,,,,,,,,,,30.9,,69.1,,,10.0,,,,,,,,,,,,,,,,,,,63.9,,,,,,,
36
+ glm-4.6,56.0,54.0,,,,93.9,,,,,,59.8,,,,,,,,45.1,,49.5,,,,,,,,,,,,,,,,,,,,,,29.2,,,,71.9,,,81.0,81.0,,,,,17.2,17.2,30.4,89.2,,87.7,,43.0,,73.5,,,,70.0,,,,,,,,,,,,,83.2,,,,,,,,,,30.0,,,,,,,,,,,,38.0,,,,,,53.8,,68.0,,,,,71.0,,,,24.5,,23.6,,,,,,,,,70.0
37
+ glm-4.7,,,,,,95.7,,,,,,,,,,,,,,52.0,,66.6,,,,,,1443.0,,,,,,,,,23.5,,,,,,,,,2.4,,,,,85.7,85.7,,,,,24.8,42.8,42.8,97.1,,93.5,,,,82.0,,,57.3,84.9,,,,,,,,,,52.0,,,84.3,,,,,,,,,,,,,,,,,,,,,,,,,,,,66.7,,73.8,,,,,,,,,41.0,,33.3,23.8,,,2376.82,,,,,
38
+ glm-5,,,,,,,95.4,,44.7,4.9,,,,,,,,,,62.0,,72.7,,,,,,,,,,,,,,,48.3,,,,,,,,,,,,,,86.0,,,,,,30.5,,50.4,,82.8,96.9,,,,82.5,,,,,,,,,,,,,,69.2,,,,,,,,,,,,,,,,,35.9,,,,,,,,,,,,,,73.3,55.1,77.8,,,,,,69.2,,,56.2,,,38.0,,,4432.12,,,,,
39
+ glm-5.1,,,,,,,95.3,,,,,,,,,,,,,79.3,,,,,,,,,75.0,,,,,,,,68.7,,,,,,,,,,,,1535.0,,86.2,86.2,,,,,34.7,52.3,52.3,94.0,89.4,94.0,,,,83.8,,,,,,,,,,11.5,,,,71.8,,,86.0,,,,,,,,,,,,,,42.7,,,,,,,,,,,38.1,,,73.3,58.4,,,,,,,70.6,,,63.5,,,40.7,,,5634.41,,,,,
40
+ gpt-3.5-turbo,,,,,,,,38.7,,,,,,,,,66.6,,,,,,,,,,,,,,,,,,,,,,70.2,,,,,,,,,,,,30.8,30.8,,,74.3,,,,,,,,68.0,,66.9,,,,,,,,,41.2,,,,0.0,,,,56.3,,,0.0,,,,,,,,,,,,,,,,,45.0,,69.1,,,,,,,,,,,,,,,,,,,,,,62.9,,,,,,,
41
+ gpt-4.1,,,52.9,51.6,48.1,46.4,,,5.5,0.4,61.5,,,,,,,,,,85.9,,,14.55,,87.9,56.7,,,,,,,65.8,65.5,,,,,44.3,,6.7,,,,,,,,,66.3,66.3,61.7,58.0,,,5.4,5.4,,28.9,,,,,87.4,,,49.1,,44.7,,0.7,1.1,,92.4,,,72.2,,,,,81.8,87.3,74.8,60.3,,,57.2,,48.1,70.8,,38.3,46.2,,,,,,,,,,,,42.3,,,,,,,,34000.0,56.0,74.0,34.0,,49.4,68.0,,30.3,,,,,,,78.7,60.9,,
42
+ gpt-4.1-mini,,,31.6,34.7,49.6,40.2,,,3.5,0.0,28.2,,,,,,,48.9,,,89.0,,,,,88.4,56.8,,,,,,,54.6,49.3,,,,,42.3,,10.9,,,,,,,,,65.0,65.0,61.7,60.5,,,3.7,3.7,,35.0,,,,,84.1,,,45.1,,,,1.1,1.8,,,,,73.1,,,,,,78.5,72.7,58.9,,,47.2,,49.6,67.0,,35.8,42.2,,,,,,,,,,,,,,,,,,,,31000.0,51.0,66.0,44.0,,36.0,55.8,,,,,,,,,68.4,55.1,,
43
+ gpt-4.1-nano,,,6.2,9.8,29.4,,,,0.0,0.0,10.7,,,,,,,,,,89.4,,,,,73.9,40.5,,,,,,,42.5,5.7,,,,,26.5,,,,,,,,,,,50.3,50.3,25.0,9.4,,,,,,,,,,,74.5,,,31.6,,,,,,,,,,56.2,,,,,,66.9,55.4,33.0,,,36.6,,29.4,57.2,,15.0,31.1,,,,,,,,,,,,,,,,,,,,9000.0,14.0,21.5,12.1,,14.0,22.6,,,,,,,,,55.2,30.2,,
44
+ gpt-4.5,,,44.9,,36.7,,,,10.3,0.8,51.4,,,,,,,,,,,,,,,90.0,55.4,,,,,,,72.3,63.0,,,,,,,,,,,,,,,,69.5,69.5,72.3,72.6,97.0,,,,,,,,,,88.2,,,54.0,,,,,,,,,,72.3,,,,,,85.1,75.2,,,,38.5,,36.7,70.8,,43.8,50.1,,,,,,,,,,,,,,,,,,38.0,37.3,,,,,,50.0,68.4,,,,,,,,,,,,
45
+ gpt-4o,,,18.2,30.7,13.1,,,,4.5,0.0,95.4,,,,,,,,,,,,,,,85.3,52.7,,,,,,,50.2,66.5,,,,83.4,35.2,,,,,,,,,,,46.0,70.1,41.7,35.4,,,5.3,5.3,,,,,91.5,,81.0,,,29.2,,31.4,,,,76.6,76.4,,,61.4,,,,90.5,75.8,81.4,68.7,59.9,,,31.9,,13.1,60.9,,27.8,39.9,79.8,,,,,,,,,,,38.2,,,,,,30.7,32.6,,,,,,42.8,60.3,,,,,,,,,,61.2,,
46
+ gpt-4o-0513,,,16.0,,9.3,,,,,,75.6,,,,58.6,,,,53.1,,,,,,76.0,,,,59.3,87.9,,10.8,759.0,,,,,,83.7,,,,,,80.5,,,,,,49.9,59.1,,,,,,,,,,,80.5,,84.3,,94.0,,57.6,33.4,48.1,,,70.2,74.6,,,,89.9,,,90.4,72.6,,,,,,,,,,,,,,,,,,,,92.2,,,,38.2,,,,,,38.8,,,,,,,,,,,,,,,,,,,,
47
+ gpt-4o-mini,,,2.7,3.6,8.6,,,49.7,,0.0,76.2,,,,,,65.9,,,,,,,,,76.6,36.8,,,,,,,52.7,38.6,,,,79.3,,,,,,,,,,,,40.2,40.2,29.0,12.6,83.0,,,,,,,,86.2,,78.4,,,27.2,48.1,,,,,67.9,73.0,,,56.5,,,,86.5,63.4,70.5,56.3,,,,24.5,,,57.9,,20.3,25.6,,,,,,,39.0,,84.9,,,9.9,,,,,,,,,,,,,22.0,44.0,,,,,64.8,,,,,,,
48
+ gpt-5,69.0,76.0,88.0,88.0,,94.6,,,65.7,9.9,,73.0,,,,,,,,54.9,90.0,63.0,91.67,21.82,,,81.1,,,,90.0,,2537.0,99.0,,5.7,,,,65.7,,2.8,,63.9,,26.3,26.3,76.4,,,85.7,88.1,78.3,73.3,,67.2,24.8,24.8,,88.3,,89.2,,73.0,,76.0,,64.0,71.3,84.5,,1.0,1.2,84.7,99.4,1.04,,,,,50.9,,87.5,,84.2,78.4,,,,,,,,69.6,69.6,,,56.2,,,,,,,42.9,,55.0,,91.98,,55.3,41.78,74.9,,112000.0,62.6,81.1,96.7,,,,35.2,41.3,30.5,29.0,,,,,86.7,84.6,,77.8
49
+ gpt-5-mini,,,71.6,,,91.1,,,,,,,,,,,,,,,89.4,,,,,,75.5,,,,,,,98.5,,,,,,62.9,,3.5,,,,22.1,22.1,,,,82.3,82.3,73.4,64.3,,,16.7,16.7,,87.8,,,,,,,,65.8,,,,0.7,1.3,,,,,,,,,,,,81.6,74.1,,,84.3,,,,,,62.3,,,,,,,,,,,,,,,,,,71.0,,75000.0,60.0,78.3,74.1,,,,,,,,,,,,78.5,82.5,,
50
+ gpt-5-nano,,,48.4,,,85.2,,,,,,,,,,,,,,,80.4,,,,,,62.7,,,,,,,96.9,,,,,,50.1,,7.3,,,,9.6,9.6,,,,71.2,71.2,64.0,43.8,,,8.7,8.7,,75.6,,,,,,,,56.1,,,,1.0,2.8,,,,,,,,,,,,75.6,62.6,,,43.2,,,,,,54.9,,,,,,,,,,,,,,,,,,54.7,,49000.0,41.0,62.3,35.5,,,,,,,,,,,,65.7,66.8,,
51
+ gpt-5.1,,75.0,,,,94.0,,,72.8,17.6,,,,,,,,,,50.8,90.0,,93.33,25.45,,,67.0,1439.0,,,91.88,,,,,,,,,,,,,,,31.0,12.5,,,,88.1,88.1,76.8,71.5,,,25.7,,,96.3,,91.67,,,,,,,,87.0,,,,,,1.04,,,,44.5,,,87.0,89.5,85.4,,,,,,,,,,,,,,,,,,,,,,,,91.04,,,50.8,76.3,69.7,,67.0,77.9,95.6,,,,47.6,,43.0,36.1,,,,,,82.9,,
52
+ gpt-5.2,,72.3,91.1,,,100.0,96.7,,86.2,52.9,,71.1,37.4,86.0,,,,,,65.8,92.0,76.1,98.33,27.27,,93.8,82.1,1438.0,,,91.25,,3148.0,96.9,,11.6,,71.3,,59.8,,91.9,59.5,,84.0,40.3,14.6,,1462.0,91.2,92.4,93.2,94.0,89.0,,63.3,34.5,45.5,45.5,99.4,,97.1,,75.4,94.8,86.3,,,,87.7,54.5,99.7,99.2,,,13.54,83.0,83.1,,60.6,57.5,,86.7,89.6,86.7,79.5,,83.8,,89.4,,,47.7,57.9,,,49.3,,0.14,0.143,47.3,,,,52.0,86.3,,38.0,91.98,67.9,72.0,55.6,80.0,74.6,,,82.0,98.7,,,,62.2,,,46.3,,,3952.0,73.1,86.0,85.9,76.8,
53
+ gpt-5.4,,,,,,,98.7,,93.7,73.3,,,49.7,,,,,,,82.7,,,,25.45,,,82.8,,76.8,,,,3168.0,,,,79.0,63.7,,,,,56.0,,,47.6,27.1,,1674.0,,92.8,92.8,93.0,89.8,,,39.8,39.8,,,97.7,95.8,,,,91.4,,,,,,,,,,54.1,92.0,,,70.6,62.5,,87.5,,,81.2,,,,,,,,,,,41.3,,0.109,89.1,75.0,,,,56.6,,,45.3,,,,57.7,,,,,,92.8,72.9,,,75.1,,,54.6,,,6144.18,,,,,
54
+ gpt-5.5,,,,,,,,,95.0,85.0,,,,,,,,,,84.4,,,,,,,,,,,,,,,,,81.8,,,,,,60.0,,,51.7,35.4,,,,93.6,93.6,,,,,41.4,57.2,,,,,,,,,,,,,,,,,,,,,,75.3,,,,,,81.2,,,,74.0,,,,,,,,,,,78.7,,,,,,,,,,,58.6,,,,,,98.0,,,,82.7,,,55.6,,,,,,,,
55
+ gpt-oss-120b,,,44.4,,96.6,97.9,,,,,,,,,,,,,,,,,,3.64,,,,,,,,,2622.0,,,,,,,,,,,,,,,,,,80.9,80.9,,,,57.6,19.0,14.9,,,,,,69.5,,,,,,81.9,,,,,,,,,,,,,,81.3,,,,,,,,,,,,,,,,,,,,,,,,,,,,,62.4,,,,,,,49.2,67.8,18.7,,,,,,,,,,,
56
+ grok-3-beta,,,,,52.2,77.3,,,31.9,0.0,,,,,,,,,,,,,,,,,,1402.0,,,,,,33.6,,,,,,,,,,,,,,,,,80.2,84.6,,,,,18.2,,,,,,87.3,,83.4,,,,,57.0,,,,,99.2,,,,,,,,79.9,,73.2,,,,,,93.3,,,,,,,,,,,,,,,,43.6,,,62.8,,,48.5,,,,,,,,,,17.5,,,,4.76,,,,,,
57
+ grok-4,73.0,68.0,,,,91.7,,,66.7,15.9,,,,,,,,,,,,,95.0,,,,,1465.0,,,83.75,,,,,,,,,,,,,,,13.0,,,,,87.5,88.4,,,,,25.4,50.7,,90.0,,88.33,,,,73.1,,,,79.0,,,,,,2.08,,,,,,,87.0,,,,,,,,,,,,,,,,,,,,,,,,,,85.85,,,46.5,,,,,,,,,,23.1,39.0,,,,37.5,,,,,,
58
+ grok-4.1,,,,,,91.9,,,,,,,,,,,,,,,,,97.5,,,,,,,,84.38,,,,,,,,,,,97.0,,,,,,,,85.6,84.3,,,,,,17.6,,,,,93.33,,,,,,,60.0,,,,,,,5.21,,,,,,,,86.8,,63.0,,54.6,,,,,,,,,,,,,,,,,,,,19.5,84.6,,,,50.6,,,,,,,,,,,,,,,1107.0,,,,,
59
+ kimi-k2,50.0,52.0,60.0,60.0,69.6,49.5,,,,,54.5,54.2,,,,,,,,14.1,,28.8,,,,,,,,,,74.3,,,,,,,,,,,,29.5,,,,60.2,,,75.1,75.8,,,97.3,43.8,4.7,4.7,,38.8,,,,42.0,89.8,45.8,,,76.4,53.7,,,,89.1,97.4,,,,,,,,81.1,,,,,,,,72.0,,33.5,54.1,,85.7,,25.5,,,,,,,30.7,,31.0,,,57.2,47.3,27.67,65.8,,,56.5,70.6,65.8,,,,44.5,,23.0,,,,,,,,,61.0
60
+ kimi-k2.5,,70.0,,,,96.1,94.5,,65.3,11.8,,,36.5,,,,,,,60.6,,62.3,98.33,47.27,,,77.5,,,,91.25,,,,,,41.3,77.1,,,,,,,,,,,,89.3,87.6,87.6,,,,,30.1,50.2,50.2,95.4,81.3,91.1,,70.2,93.9,81.8,,,,85.0,61.0,,,,,8.85,84.2,,,63.8,29.5,,87.1,86.0,84.3,78.5,,,,,,,,62.7,,,32.0,57.4,0.112,0.112,63.3,,,,48.7,,,,90.57,69.2,73.0,50.7,76.8,,,,,,66.0,,,50.8,,,27.8,,,1198.46,,87.4,86.6,72.7,
61
+ kimi-k2.6,,,,,,,96.4,,,,,,39.8,,,,,,,83.2,,,,,,,80.4,,75.9,,,,,,,,,83.0,,,,,,,,,,,1482.0,,90.5,90.5,,,,,36.4,36.4,,,92.7,,,,,86.0,,,,89.6,,,,,,24.0,87.4,,,66.6,55.9,,87.1,,,79.4,,,,,,,,,,,,60.6,,,73.1,,,,52.2,,,36.9,,,76.7,58.6,80.2,,,,,,,,,66.7,,,50.0,,,,,,,80.8,
62
+ llama-3.1-405b-instruct,,,5.8,,23.3,,,,,,,,,,88.5,,,46.2,59.4,,,,,,61.5,,,,50.4,84.7,,6.8,,,,,,,88.7,,,,,,70.0,,,,,,51.1,49.0,,,96.8,,,,,,,,77.2,,86.0,,93.9,,,28.4,36.1,,,73.8,73.8,,,,88.6,,,,73.3,,,,,,,,,,,,,,,,,,,,90.4,,,,17.1,,,,,,24.5,,,,,,,,,,,,,,,,,,,,
63
+ llama-3.1-8b,,,,,,,,25.8,,,,,,,50.9,,71.9,,41.9,,,,,,,,,,,,,,,,,,,,61.7,,,,,,,,,,,,,23.4,,,83.4,,,,,,,,72.6,,80.6,,90.1,,,,,,,42.5,,,,,72.8,,,,46.5,,,,,,,,,,,,,,,,,,,28.4,73.6,70.2,,,,,,,,,,,,,,,,,,,,,,55.1,,,,,,,
64
+ llama-3.2-1b-instruct,,,,,,0.33,,10.1,,,,,,,,21.44,40.2,,,,,,,,,,,,,,,,,,,,,,32.2,,,,,,,,,,,,16.57,,,,39.04,,,,,,,,,15.93,52.37,,,,,,,,,21.6,23.4,,,,,,,,20.8,,,,,,,,,30.16,,,,,,,,,,13.8,,87.2,,,,,,,,,,,,,,,,,,,,,,41.5,,,,,,,
65
+ llama-3.3-70b-instruct,,,,,,,,36.5,,,65.5,,,,51.4,,85.8,46.9,58.0,,,,,,,,,,,,,,,,,,,,90.2,,,,,,,,,,,,49.1,50.5,,,93.6,,,,,,,,78.91,,89.3,,92.8,,57.6,32.9,,,,71.8,66.31,,,,84.4,,,89.1,64.4,,,,,,,,,,,,,,,,,,,48.2,85.6,70.4,,,20.9,,,,,,29.4,,,,,,,,,,,,,66.1,,,,,,,
66
+ llama-4-maverick,,,,,,,,,4.4,0.0,91.8,,,,,,,49.7,,,,,,54.55,,,,1417.0,,,,,,,,0.0,,,,,,,,,,,,,,,69.8,69.8,,,,,,,,,,,85.4,,88.9,,,,,43.4,,,,61.2,90.0,,,73.7,,,,92.3,80.5,,73.4,59.6,,,,,,,,,,76.4,,,,,,,,,,,,,,,,,46.5,,,,,,,,,,15.5,,,,,,,,,,
67
+ minimax-m2,61.0,61.0,,,,78.3,,,,,,66.8,,,,,,,,44.0,,48.5,,,,,,1346.0,,,,,,,,,,,,,,,,65.5,,,,75.7,,,77.7,78.0,,,,,12.5,12.5,31.8,,,,,72.0,,,,,,83.0,,,,,,,,,,,24.4,,82.0,,,,,,,,,,36.2,,,,,,,,,,,,36.0,,,,,,56.5,,69.4,,,,,87.0,,,,30.0,42.0,24.0,16.0,,,,,,,,72.0
68
+ minimax-m2.1,,62.0,,,,83.0,,,,,,,,,,,,,,62.0,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,83.0,81.0,,,,,22.2,22.0,,,,,,70.0,,,,,,78.0,,,,,,,,,,,,,88.0,,,,,,,,,,49.4,,,,,,,,,,,,41.0,,,,,,72.5,,67.0,,,,,,,,,29.2,,,43.5,,,,,,,,
69
+ ministral-8b,,,,,86.0,78.7,,31.4,,,,,,,51.8,,70.8,,33.2,,,,,,,,,,,,,,,,,,,,56.2,,,,,,,,,,,,66.8,23.4,,,80.0,,,,,,,,34.8,,56.4,,88.3,,,61.6,,,,40.0,,,,,61.1,,,,43.0,,,,,,,,86.0,,,,,,,,,,,20.2,62.0,56.2,,,,,,,,,,,,,,,,,,,,,,55.5,,,,,,,
70
+ muse-spark,,,,,,,,,,42.5,,,,,,,,,,,,,,,,,86.4,,,,,,,,,,,,,64.7,,,,,,,,,1444.0,,89.5,89.5,,,,,42.8,58.4,50.4,,,,,,,,,,,,,,,,,,,,,,,,,,,80.4,,,,,,,,,,,,,,,,,,,,84.1,,,,,,52.4,77.4,,,,,91.5,,,,59.0,,,,,,,,,,,
71
+ o1-high,,,61.7,64.6,74.3,71.4,,,,,,,,,,,,,,,,,,,,88.9,55.1,,,,,,1891.0,95.3,47.6,,,,,,,,,,,5.5,5.5,,,,75.7,78.0,62.0,50.9,97.1,,,,,48.3,,,88.1,,92.2,,,51.3,67.0,63.4,,,,96.4,,,,71.8,,,,89.3,,87.7,77.6,,,,22.1,,74.3,77.9,,44.9,52.9,,,,,,,,,,,,47.0,,,,,,48.9,,,,,,,50.0,70.8,,,,,,,,,,,,
72
+ o3-high,,69.0,79.6,81.3,,88.9,,,60.8,6.5,,,,,,,,,,,88.3,,95.83,,,,78.6,,,,79.38,,,98.4,,1.4,,,,64.0,,23.5,,,,15.8,15.8,,,,83.3,83.3,77.3,72.9,,,20.2,14.7,,81.7,,,88.4,69.3,,,,47.4,,,,5.2,6.8,,98.1,,,86.8,,,,,,,82.9,76.4,,,,,91.6,,,60.4,60.4,,,,,,,,,,,,48.6,,87.74,,,,,,86000.0,64.8,80.2,58.2,,,,,30.2,,,,,,,84.9,83.3,,
73
+ o3-mini-high,,,60.4,66.7,87.3,86.5,,,25.8,3.0,81.9,,,63.6,,,,,,,,,,,,,,1363.0,,,,,2130.0,98.7,17.6,,,,,,,,,,,12.4,,,,,77.2,,51.0,58.3,,,,,,67.5,,,97.6,,93.9,,,50.0,,68.8,,,,,97.9,,,,,,,,82.4,80.7,,,,,18.7,,,79.5,,39.9,50.2,,,,,,,,,,,,13.8,,,52.2,,,61.0,,,,,,,32.4,57.6,,,,,,2.08,,,,,,
74
+ o4-mini-high,,,58.2,68.9,,92.7,,,58.7,6.1,,,,,,,,,,51.5,80.0,,86.67,,,,72.0,,,,84.38,,,96.1,,0.6,,,,56.5,,38.7,,,,15.4,,,,,81.4,81.4,62.3,51.1,,,14.7,14.7,,85.0,,,99.3,,92.4,,,44.7,,,,3.0,8.9,,98.2,,,84.3,,,,,,,81.6,73.4,,,,,93.4,,,43.0,57.5,,,,,,,,,,,,19.3,,88.68,,,,,,66000.0,60.2,70.5,40.5,,49.2,71.8,,18.5,,,,,,,79.5,79.4,,
75
+ qwen-2.5-14b-instruct,,,,,,,,34.6,,,70.2,,,,,,78.4,,,,,,,,,,,,,,,,,,,,,,85.5,,,,,,,,,,,,42.9,45.5,,,94.8,,,,,,,,72.1,,78.7,,,,46.6,,,,,80.0,75.6,,,,63.2,,,79.6,63.2,,,,,,,,,,,,,72.8,,,,,,21.1,,79.3,,,5.4,,,,,,,,,,,,,,,,,,,58.4,,,,,,,
76
+ qwen-2.5-32b,,,,,,,,39.1,,,,,,,,,82.3,,,,,,,,,,,,,,,,,,,,,,48.3,,,,,,,,,,,,49.5,49.5,,,87.5,,,,,,,,88.4,,82.4,,,,,,,,,77.9,,,,,67.2,,,,69.0,,,,,,,,,,,,,75.4,,,,,,26.1,,82.4,,,,,,,,,,,,,,,,,,,,,,70.6,,,,,,,
77
+ qwen-2.5-7b,,,,,,,,29.7,,,52.0,,,,,,70.2,,,,,,,,,,,,,,,,,,,,,,54.4,,,,,,,,,,,,36.4,36.4,,,83.8,,,,,,,,84.8,,74.7,,,,35.9,28.7,,,,69.9,,,,,,,,,56.3,,,,,,,,,,,,,70.4,,,,,,18.1,,75.0,,,,,,,,,,,,,,,,,,,,,,63.1,,,,,,,
78
+ qwen2.5-72b-instruct,,,7.6,,23.3,,,47.7,,,78.4,,,,,,80.4,45.8,50.1,,,,,,86.1,,,,48.4,91.4,,15.9,,,,,,,76.7,,,,,,69.8,,,,,,49.0,,,,89.5,,,,,,,,77.3,,84.1,,,,55.3,31.1,39.4,,,75.9,80.0,,,,88.6,,,87.3,71.6,,,,,,,,,,,,,,,,,,,30.6,83.2,87.0,,,9.1,,,,,,23.8,,,,,,,,,,,,,69.9,,,,,,,
79
+ qwen3-235b,,67.0,57.3,57.3,85.7,81.5,,,11.0,1.3,95.6,,,,70.8,70.8,,,,,,,,,,,,,,,,,2056.0,,,,,,,,,,,,,,,,,,71.1,81.1,,,,,,18.2,,62.5,,,90.0,,83.4,,,,77.1,70.7,,,,71.8,98.0,5.21,,,,,,83.5,,84.3,,,,,,,80.8,71.9,,,,87.9,,32.5,,,,,,,,,13.2,,,64.9,,,69.6,,,,,,,46.0,67.8,,6.6,,,,,,,,,,
80
+ qwen3-30b-a3b,,,,,80.4,70.9,,,,,91.0,,,,69.1,69.1,,,,,,,,,,,,,,,,,1974.0,,,,,,,,,,,,,,,,,,65.8,65.8,,,,,,,,,,,,,86.5,,,,74.3,62.6,,,,,98.0,,,,,,,,,78.4,,,,,,,73.9,72.2,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
81
+ qwen3-32b,,,,,81.4,72.9,,,,,93.8,,,,70.3,70.3,,,,,,,,,,,,,,,,,1977.0,,,,,,,,,,,,,,,,,,68.4,,,,,,,,,,,,85.0,37.3,85.0,,,,74.9,65.7,,,,,97.2,,,,,,,,,80.6,,,,,,,75.0,73.0,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,15.5,,,,,,,,,,
82
+ qwen3-8b,,,,,76.0,67.3,,,,,85.8,,,,,68.1,,,,,,,,,,,,,,,,,1785.0,,,,,,,,,,,,,,,,,,62.0,,,,,,,,,,,,82.0,35.0,85.0,,,,67.1,57.5,,,,,97.4,,,,,,,,,74.4,,,,,,,65.4,71.2,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
83
+ qwen3.5-397b,,68.7,,,,,91.3,,,,,,,,,,,,,69.0,,70.3,,65.45,,,80.8,,,,,,,,,,,,,,,,,,,,,,,89.8,88.4,,,,,,28.7,,,94.8,87.88,92.7,,76.5,92.6,80.9,,,,83.6,63.2,,,,,,88.6,,,,46.1,,87.8,88.5,85.0,79.0,,,,,,,,67.6,,,,,,,,,,,,65.6,,,,70.4,69.3,,76.4,,,,,,,,,52.5,,,38.3,,,,,87.5,84.7,,
84
+ qwen3.6-plus,,68.3,,,,,95.1,,,,,,,,,,,,,,,,,,93.3,,,,,,,,,,,,,,,65.7,,,,,,,,,,89.8,90.4,90.4,,,,,28.8,28.8,,96.7,87.8,94.6,,74.2,94.3,83.8,,,,87.1,62.0,,,,,,88.0,,,74.1,48.2,,88.5,89.5,86.0,78.8,,,,,,,,,,,37.9,,,91.2,,,,,,68.2,,,,71.6,73.8,56.6,78.8,,,,,,70.7,,,61.6,,,39.8,,,5114.87,,84.2,84.0,74.3,
85
+ qwq-32b,,,,,79.5,69.5,,82.4,,,89.5,,,,66.4,66.4,89.6,44.6,,,,,,,,,,,,,,,1982.0,,,,,,54.7,,,,,,,,,,,,65.6,65.2,,,95.5,,,,,47.5,,,,,83.9,,,,72.0,62.7,,,,98.1,98.0,,,,,,,,,80.9,,,,,,,68.0,68.3,,,,,,,,,,,,69.9,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
data/scores_all.csv ADDED
The diff for this file is too large to render. See raw diff
 
data/scores_all.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:da24501af1071c34d4e75fc259fd891d7e3cc0998690ed71c849001c075648d1
3
+ size 172183
data/scores_paper.csv ADDED
The diff for this file is too large to render. See raw diff
 
data/scores_paper.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:254749e5220c659580343d6d39ea2ab334cdcefe68131671ce912e1ee26d05f9
3
+ size 96434
metadata.json ADDED
@@ -0,0 +1,433 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "repo_id": "yzeng58/benchpress-score-matrix",
3
+ "generated_at_utc": "2026-05-04T20:15:44.377454+00:00",
4
+ "source_git_commit": "67d07b70a5fc5c1bca666f85b235502bb6b6577b",
5
+ "source_git_branch": "yz",
6
+ "source_data_file": "benchpress/data/llm_benchmark_data.json",
7
+ "source_data_dirty": true,
8
+ "n_models_all": 189,
9
+ "n_benchmarks_all": 316,
10
+ "n_scores_all": 4903,
11
+ "paper_matrix": {
12
+ "m_threshold": 15,
13
+ "b_threshold": 8,
14
+ "iterations": 5,
15
+ "n_models": 84,
16
+ "n_benchmarks": 133,
17
+ "n_observations": 2604,
18
+ "fill_rate": 0.23308270676691728,
19
+ "dropped_models": [
20
+ "amazon-nova-premier",
21
+ "amazon-nova-pro",
22
+ "claude-3-haiku",
23
+ "claude-3-opus",
24
+ "claude-3-sonnet",
25
+ "claude-3.5-haiku",
26
+ "claude-mythos",
27
+ "claude-sonnet-3.7",
28
+ "codestral-22b",
29
+ "command-r7b",
30
+ "deepseek-3.1",
31
+ "deepseek-r1-distill-llama-8b",
32
+ "deepseek-r1-distill-qwen-1.5b",
33
+ "deepseek-r1-distill-qwen-14b",
34
+ "deepseek-r1-distill-qwen-32b",
35
+ "deepseek-r1-distill-qwen-7b",
36
+ "deepseek-v3-0324",
37
+ "devstral-2",
38
+ "exaone-4.0-32b",
39
+ "exaone-deep-32b",
40
+ "falcon3-10b",
41
+ "gemini-1.5-flash-8b",
42
+ "gemini-2.0-flash-lite",
43
+ "gemini-2.0-pro",
44
+ "gemini-diffusion",
45
+ "gemma-2-2b",
46
+ "gemma-3-12b",
47
+ "gemma-3-4b",
48
+ "gemma-3n-e2b-instructed",
49
+ "gemma-3n-e2b-instructed-litert-preview",
50
+ "gemma-3n-e4b-instructed",
51
+ "gemma-3n-e4b-instructed-litert-preview",
52
+ "gemma-4-26b-a4b",
53
+ "gemma-4-31b",
54
+ "gemma-4-e2b",
55
+ "gemma-4-e4b",
56
+ "gpt-4-turbo",
57
+ "gpt-5.2-codex",
58
+ "gpt-5.3-codex",
59
+ "gpt-oss-20b",
60
+ "granite-3.3-8b-instruct",
61
+ "granite-4.0-1b",
62
+ "granite-4.0-h-1b",
63
+ "grok-1.5",
64
+ "grok-2",
65
+ "grok-2-mini",
66
+ "grok-4.20",
67
+ "ibm-granite-4.0-tiny-preview",
68
+ "internlm3-8b",
69
+ "lfm2.5-1.2b-instruct",
70
+ "llama-3.1-70b",
71
+ "llama-4-behemoth",
72
+ "llama-4-scout",
73
+ "longcat-flash-chat",
74
+ "minicpm-sala",
75
+ "minimax-m2.5",
76
+ "minimax-m2.7",
77
+ "mistral-large-2",
78
+ "mistral-large-3",
79
+ "mistral-medium-3",
80
+ "mistral-small-24b",
81
+ "mistral-small-3-24b-instruct",
82
+ "mistral-small-3.1",
83
+ "nemotron-ultra-253b",
84
+ "nova-lite",
85
+ "nova-micro",
86
+ "o1-mini",
87
+ "olmo-1b",
88
+ "olmo-2-13b",
89
+ "olmo-2-1b",
90
+ "olmo-2-32b",
91
+ "olmo-2-7b",
92
+ "olmo-7b-0424",
93
+ "openthinker2-32b",
94
+ "phi-3-14b",
95
+ "phi-3.5-mini-instruct",
96
+ "phi-3.5-moe-instruct",
97
+ "phi-4",
98
+ "phi-4-mini",
99
+ "phi-4-reasoning",
100
+ "phi-4-reasoning-plus",
101
+ "pixtral-12b",
102
+ "qwen-2.5-1.5b",
103
+ "qwen2-72b-instruct",
104
+ "qwen2-7b-instruct",
105
+ "qwen2.5-coder-32b-instruct",
106
+ "qwen2.5-coder-7b-instruct",
107
+ "qwen2.5-omni-7b",
108
+ "qwen2.5-vl-32b-instruct",
109
+ "qwen3-0.6b",
110
+ "qwen3-1.7b",
111
+ "qwen3-14b",
112
+ "qwen3-4b",
113
+ "sarvam-30b",
114
+ "seed-thinking-v1.5",
115
+ "smollm2-1.7b",
116
+ "tulu-3-8b"
117
+ ],
118
+ "dropped_benchmarks": [
119
+ "aethercode",
120
+ "agentcompany",
121
+ "ai2d",
122
+ "aider_polyglot",
123
+ "ainstein_bench",
124
+ "all_angles",
125
+ "apex_agents",
126
+ "arc_challenge",
127
+ "arcagi1_image",
128
+ "arcagi2_image",
129
+ "babe",
130
+ "bfcl_v3_multiturn",
131
+ "bfcl_v4",
132
+ "bigbench_extra_hard",
133
+ "biobench",
134
+ "bixbench",
135
+ "blink",
136
+ "browsecomp_cm",
137
+ "browsecomp_long_context_256k",
138
+ "cgbench",
139
+ "chartqa",
140
+ "chartqapro",
141
+ "cl_bench",
142
+ "claw_eval_pass3",
143
+ "cmmlu",
144
+ "codesimpleqa",
145
+ "contphy",
146
+ "corpusqa_1m",
147
+ "countbench",
148
+ "covost2",
149
+ "creative_writing_v3",
150
+ "crossvid",
151
+ "ctf_internal",
152
+ "cybench",
153
+ "cybersecurity_ctf",
154
+ "da_2k",
155
+ "deepconsult",
156
+ "deepresearchbench",
157
+ "deepsearchqa_f1",
158
+ "der2_bench",
159
+ "disco_x",
160
+ "docvqa",
161
+ "dude",
162
+ "dynamath",
163
+ "egoschema",
164
+ "egotempo",
165
+ "emma",
166
+ "encyclo_k",
167
+ "eq_bench3",
168
+ "expert_swe",
169
+ "facts_benchmark",
170
+ "finsearchcomp",
171
+ "finsearchcompt23",
172
+ "flenqa_3k",
173
+ "fleurs",
174
+ "frontier_science_research",
175
+ "frontiersci_olympiad",
176
+ "frontiersci_research",
177
+ "fsc_147",
178
+ "gdpval_diamond",
179
+ "gdpval_oai_woe",
180
+ "genebench",
181
+ "graphwalks_bfs_128k_plus",
182
+ "graphwalks_bfs_256k_1m",
183
+ "graphwalks_parents_128k_plus",
184
+ "graphwalks_parents_256k_1m",
185
+ "hallusionbench",
186
+ "healthbench_consensus",
187
+ "hellaswag",
188
+ "hiddenmath",
189
+ "hipho",
190
+ "hle_verified",
191
+ "hle_vl",
192
+ "humaneval_plus",
193
+ "ib_modeling",
194
+ "imo_2025",
195
+ "infovqa",
196
+ "inverse_ifeval",
197
+ "korbench",
198
+ "livesports_3k",
199
+ "loft_128k",
200
+ "logicvista",
201
+ "longdocurl",
202
+ "longform_writing",
203
+ "longvideobench",
204
+ "lpfqa",
205
+ "lvbench",
206
+ "mars_bench",
207
+ "mathcanvas",
208
+ "mathkangaroo",
209
+ "mathvista_mini",
210
+ "medxpertqa_mm",
211
+ "medxpertqa_text",
212
+ "minedojo_verified",
213
+ "minerva",
214
+ "mm_browsecomp",
215
+ "mme_cc",
216
+ "mmlongbench",
217
+ "mmlongbench_doc",
218
+ "mmsibench_circular",
219
+ "mmstar",
220
+ "mmvu",
221
+ "morse_500",
222
+ "motionbench",
223
+ "mrcr_v2_2needle_1m",
224
+ "mrcr_v2_2needle_256k",
225
+ "mt_bench_101",
226
+ "mtvqa",
227
+ "muirbench",
228
+ "multilingual_mmlu",
229
+ "nl2repo_pass1",
230
+ "ntrex",
231
+ "ocrbench",
232
+ "ocrbench_v2",
233
+ "odvbench",
234
+ "officeqa",
235
+ "officeqa_pro",
236
+ "omnimath",
237
+ "ovbench",
238
+ "ovobench",
239
+ "paperbench",
240
+ "phibench",
241
+ "phybench",
242
+ "phyx_openended",
243
+ "point_bench",
244
+ "procbench",
245
+ "realworldqa",
246
+ "refspatialbench",
247
+ "researchrubrics",
248
+ "ruler_128k",
249
+ "ruler_32k",
250
+ "seal_0",
251
+ "sfe",
252
+ "simplevqa",
253
+ "spreadsheetbench_verified",
254
+ "superchem",
255
+ "swe_bench_multimodal",
256
+ "swe_evo",
257
+ "tau_bench_telecom",
258
+ "tempcompass",
259
+ "tob_complex_workflows",
260
+ "tob_compositional_tasks",
261
+ "tob_information_extraction",
262
+ "tob_k12_education",
263
+ "tob_referenceqa",
264
+ "tob_text_classification",
265
+ "tomato",
266
+ "treebench",
267
+ "tvbench",
268
+ "usamo_2026",
269
+ "videoeval_pro",
270
+ "videoholmes",
271
+ "videoreasonbench",
272
+ "videosimpleqa",
273
+ "visfactor",
274
+ "vispeak",
275
+ "visulogic",
276
+ "vitabench",
277
+ "viverbench",
278
+ "vlms_are_biased",
279
+ "vlms_are_blind",
280
+ "vpct",
281
+ "wildbench",
282
+ "world_travel_text",
283
+ "world_travel_vlm",
284
+ "worldvqa",
285
+ "xlrs_macro",
286
+ "xpert_bench",
287
+ "zerobench_main",
288
+ "zerobench_sub",
289
+ "zerobench_tools"
290
+ ],
291
+ "excluded_models": [
292
+ "deepseek-v3.2-speciale",
293
+ "gemma-3-1b",
294
+ "granite-3.3-8b-base",
295
+ "kimi-k2-thinking",
296
+ "lfm2.5-1.2b-thinking",
297
+ "llama-3.2-1b",
298
+ "qwen-2.5-14b"
299
+ ],
300
+ "excluded_benchmarks": [
301
+ "apex_shortlist",
302
+ "codeforces_avg8",
303
+ "codeforces_pass8",
304
+ "global_mmlu_lite",
305
+ "healthbench_hard",
306
+ "livecodebench_pro",
307
+ "livecodebench_v5",
308
+ "livecodebench_v6",
309
+ "mmlu",
310
+ "mmlu_redux",
311
+ "tau1_bench_avg",
312
+ "tau2_bench_avg"
313
+ ],
314
+ "model_canonical_rules": [
315
+ {
316
+ "drop_id": "deepseek-v3.2-speciale",
317
+ "keep_id": "deepseek-v3.2",
318
+ "family": "DeepSeek-V3.2",
319
+ "reason": "same base model; keep the default-effort row because it has broader coverage"
320
+ },
321
+ {
322
+ "drop_id": "kimi-k2-thinking",
323
+ "keep_id": "kimi-k2",
324
+ "family": "Kimi K2",
325
+ "reason": "same base model with a different mode; keep the broader-coverage default row"
326
+ },
327
+ {
328
+ "drop_id": "lfm2.5-1.2b-thinking",
329
+ "keep_id": "lfm2.5-1.2b-instruct",
330
+ "family": "LFM2.5-1.2B",
331
+ "reason": "same base model with a different mode; keep the instruct/non-thinking row"
332
+ },
333
+ {
334
+ "drop_id": "gemma-3-1b",
335
+ "keep_id": "gemma-3-1b-it",
336
+ "family": "Gemma 3 1B",
337
+ "reason": "same base model; many sources do not disambiguate base vs instruct, so use the base-only cells to fill missing instruct cells (small differences within evaluation noise)"
338
+ },
339
+ {
340
+ "drop_id": "granite-3.3-8b-base",
341
+ "keep_id": "granite-3.3-8b-instruct",
342
+ "family": "Granite 3.3 8B",
343
+ "reason": "same base model; many sources do not disambiguate base vs instruct, so use the base-only cells to fill missing instruct cells (small differences within evaluation noise)"
344
+ },
345
+ {
346
+ "drop_id": "llama-3.2-1b",
347
+ "keep_id": "llama-3.2-1b-instruct",
348
+ "family": "Llama 3.2 1B",
349
+ "reason": "same base model; many sources do not disambiguate base vs instruct, so use the base-only cells to fill missing instruct cells (small differences within evaluation noise)"
350
+ },
351
+ {
352
+ "drop_id": "qwen-2.5-14b",
353
+ "keep_id": "qwen-2.5-14b-instruct",
354
+ "family": "Qwen2.5 14B",
355
+ "reason": "same base model; many sources do not disambiguate base vs instruct, so use the base-only cells to fill missing instruct cells (small differences within evaluation noise)"
356
+ }
357
+ ],
358
+ "benchmark_canonical_rules": [
359
+ {
360
+ "drop_id": "codeforces_avg8",
361
+ "keep_id": "codeforces_rating",
362
+ "family": "Codeforces",
363
+ "reason": "same task family; keep Codeforces Rating because it has broader coverage"
364
+ },
365
+ {
366
+ "drop_id": "codeforces_pass8",
367
+ "keep_id": "codeforces_rating",
368
+ "family": "Codeforces",
369
+ "reason": "same task family; keep Codeforces Rating because it has broader coverage"
370
+ },
371
+ {
372
+ "drop_id": "livecodebench_pro",
373
+ "keep_id": "livecodebench",
374
+ "family": "LiveCodeBench",
375
+ "reason": "same benchmark family; keep the standard LiveCodeBench because it has broader coverage"
376
+ },
377
+ {
378
+ "drop_id": "tau1_bench_avg",
379
+ "keep_id": "tau_bench_airline/tau_bench_retail",
380
+ "family": "τ-bench",
381
+ "reason": "composite average of existing domain benchmarks; keep per-domain benchmarks"
382
+ },
383
+ {
384
+ "drop_id": "tau2_bench_avg",
385
+ "keep_id": "tau2_bench_airline/tau2_bench_retail/tau2_bench_telecom",
386
+ "family": "τ²-bench",
387
+ "reason": "composite average of existing domain benchmarks; keep per-domain benchmarks"
388
+ },
389
+ {
390
+ "drop_id": "apex_shortlist",
391
+ "keep_id": "matharena_apex_2025",
392
+ "family": "MathArena Apex",
393
+ "reason": "subset/shortlist variant; keep MathArena Apex 2025 because it has broader coverage"
394
+ },
395
+ {
396
+ "drop_id": "healthbench_hard",
397
+ "keep_id": "healthbench",
398
+ "family": "HealthBench",
399
+ "reason": "hard-subset variant; keep HealthBench because it has broader coverage"
400
+ },
401
+ {
402
+ "drop_id": "mmlu",
403
+ "keep_id": "mmlu_pro",
404
+ "family": "MMLU",
405
+ "reason": "same benchmark family; keep MMLU-Pro because it has broader coverage"
406
+ },
407
+ {
408
+ "drop_id": "mmlu_redux",
409
+ "keep_id": "mmlu_pro",
410
+ "family": "MMLU",
411
+ "reason": "same benchmark family; keep MMLU-Pro because it has broader coverage"
412
+ },
413
+ {
414
+ "drop_id": "global_mmlu_lite",
415
+ "keep_id": "mmlu_pro",
416
+ "family": "MMLU",
417
+ "reason": "same benchmark family; keep MMLU-Pro because it has broader coverage"
418
+ },
419
+ {
420
+ "drop_id": "livecodebench_v5",
421
+ "keep_id": "livecodebench",
422
+ "family": "LiveCodeBench",
423
+ "reason": "same benchmark family; use v5 only to fill missing standard LiveCodeBench cells, marking them non-canonical because the version differs"
424
+ },
425
+ {
426
+ "drop_id": "livecodebench_v6",
427
+ "keep_id": "livecodebench",
428
+ "family": "LiveCodeBench",
429
+ "reason": "same benchmark family; use v6 only to fill missing standard LiveCodeBench cells, marking them non-canonical because the version differs"
430
+ }
431
+ ]
432
+ }
433
+ }