BenchLM Benchmarks

219 benchmarks · 818 model scores · Data from Sep 23, 2026

All Coding Agentic Reasoning Knowledge Multimodal Math Multilingual Instruction Following External

Reasoning23 benchmarks

0 models

3 models

3Soofi S 30B-A3BSoofi Project78.8%

4MiniCPM5-1BOpenBMB · Open weight

5Gemma 4 12BGoogle · Open weight

2 models

3Celeris-1Celeris81.4%

4Soofi S 30B-A3BSoofi Project · Open weight

0 models

0 models

0 models

135 models

1Claude Opus 4.7 (Adaptive)Anthropic5122.60

2Claude Opus 5Anthropic5068.25

3Claude Fable 5Anthropic4561.82

4Claude Opus 4.6 (Adaptive)Anthropic · Closed

5Kimi K3Moonshot AI · Closed

62 models

1Claude Fable 5Anthropic97.6%

2GPT-5.5OpenAI83.3%

3GPT-5.4OpenAI70.2%

4GPT-5.2OpenAI · Closed

5Claude Opus 4.7Anthropic · Closed

14 models

1Qwen3.8 MaxAlibaba66.3%

2Claude Opus 4.5Anthropic64.4%

3Qwen3.5 397BAlibaba63.2%

4Qwen3.6 PlusAlibaba · Closed

5Nemotron 3 UltraNVIDIA · Open weight

9 models

1Sakana Fugu-UltraSakana AI93.6%

2Qwen3.8 MaxAlibaba92.9%

3Qwen3.7 PlusAlibaba91.7%

4Qwen3.7 MaxAlibaba · Closed

5Sakana FuguSakana AI · Closed

2 models

1Gemini 3.7 FlashGoogle97%

2GPT-5.5OpenAI83.1%

2 models

1GPT-5.5OpenAI87.5%

2Claude Opus 4.7 (Adaptive)Anthropic59.2%

graphwalks Bfs128k

1 models

1MAI-Thinking-1Microsoft90%

graphwalks Parents128k

0 models

4 models

1DeepSeek V4 Pro 0813DeepSeek83.5%

3DeepSeek V4 Flash 0731DeepSeek78.7%

5Muse Spark 1.1Meta · Closed

8Gemini 3.5 FlashGoogle · Closed

2 models

1DeepSeek V4 Pro 0813DeepSeek62.0%

2DeepSeek V4 Flash 0731DeepSeek60.5%

23 models

1GPT-6 AstraOpenAI95%

2GPT-5.6 SolOpenAI92.5%

3Claude Opus 5Anthropic90.4%

4Claude Fable 5.1Anthropic · Closed

5GPT-5.5OpenAI · Closed

4 models

1Claude Opus 4.5Anthropic74%

2Qwen3.5 397BAlibaba68.7%

3Qwen3.6 PlusAlibaba68.3%

4GLM-5Z.AI · Open weight

66 models

1GPT-6 AstraOpenAI96.0%

2Sakana Fugu-UltraSakana AI95.5%

3Sakana FuguSakana AI95.5%

4GPT-5.6 SolOpenAI · Closed

5Gemini 3.1 ProGoogle · Closed

195 models

1Kimi K3Moonshot AI88.7%

2Step 5 PreviewStepFun88.3%

3MiMo-V2.6-ProXiaomi86.3%

4Claude Fable 5.1Anthropic · Closed

5Claude Opus 5.5Anthropic · Closed

198 models

1GPT-5.6 SolOpenAI32.3%

2GPT-6 AstraOpenAI31.7%

3Claude Opus 5.5Anthropic31.7%

4GPT-6 SolOpenAI · Closed

5GPT-5.5 ProOpenAI · Closed

bullshit Bench V2

96 models

1Claude Opus 4.8Anthropic95%

2Qwen3.8 MaxAlibaba95%

4Claude Sonnet 4.6Anthropic · Closed

5Claude Opus 4.5Anthropic · Closed

7Claude Opus 4.6Anthropic · Closed

0 models