BenchLM Benchmarks
207 benchmarks · 5400 model scores · Data from Jul 25, 2026
Coding24 benchmarks
humaneval
2 models
1DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
2Soofi S 30B-A3BSoofi Projectsoofi-s-30b-a3b
codeforces
3 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
3DeepSeek V4 Pro (High)DeepSeekdeepseek-v4-pro-high
4DeepSeek V4 Flash (High)DeepSeek · Open weight
swe Verified
59 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Mythos 5Anthropicclaude-mythos-5
3Claude Fable 5Anthropicclaude-fable-5
4Claude Opus 4.8Anthropic · Closed
5Claude Opus 4.7 (Adaptive)Anthropic · Closed
+54 moreswe Rebench
13 models
1Claude Opus 4.6Anthropicclaude-opus-4-6
2GLM-5Z.AIglm-5
3GLM-5.1Z.AIglm-5-1
4DeepSeek V3.2DeepSeek · Open weight
5Claude Sonnet 4.6Anthropic · Closed
+8 morelive Code Bench
6 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Qwen3.7 PlusAlibabaqwen3-7-plus
3GLM-4.7Z.AIglm-4-7
4Qwen3.6-27BAlibaba · Open weight
5Qwen3.6-35B-A3BAlibaba · Open weight
+1 morelive Code Bench V6
10 models
1Sakana Fugu-UltraSakana AIsakana-fugu-ultra
2Sakana FuguSakana AIsakana-fugu
3Kimi K2.6Moonshot AIkimi-2-6
4Qwen3.6 PlusAlibaba · Closed
5Kimi K2.5Moonshot AI · Open weight
+5 morelive Code Bench Pro
8 models
1Sakana Fugu-UltraSakana AIsakana-fugu-ultra
2Sakana FuguSakana AIsakana-fugu
3GPT-5.4OpenAIgpt-5-4
4Gemini 3.1 ProGoogle · Closed
5Muse SparkMeta · Closed
+3 moreswe Pro
54 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Claude Fable 5Anthropicclaude-fable-5
3Claude Opus 5Anthropicclaude-opus-5
4Sakana Fugu-UltraSakana AI · Closed
5Claude Opus 4.8Anthropic · Closed
+49 moreswe Multilingual
30 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3Composer 2.5Cursorcomposer-2-5
4Ornith-1.0-397BDeepReinforce AI · Open weight
5Laguna S 2.1Poolside · Open weight
+25 moreswe Multimodal
3 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3Claude Sonnet 5Anthropicclaude-sonnet-5
cursor Bench31
14 models
1Claude Fable 5Anthropicclaude-fable-5
2Claude Opus 5Anthropicclaude-opus-5
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Grok 4.5xAI · Closed
5GPT-5.6 TerraOpenAI · Closed
+9 morevibe Code Bench
74 models
3Moonshot AIOpenHandskimi/kimi-k3
4Claude Opus 4.8Anthropic82.72%
5Claude Sonnet 5Anthropic81.33%
6GPT-5.6 SolOpenAI80.50%
7Claude Opus 4.8 Claude CodeAnthropic77.48%
+69 moreprogram Bench
8 models
5Claude Opus 4.6Anthropic · Closed
7Claude Sonnet 4.6Anthropic · Closed
8GPT-5.4OpenAI · Closed
9Gemini 3.1 ProGoogle · Closed
10Gemini 3 FlashGoogle · Closed
+3 morenl2 Repo
13 models
1GLM-5.2Z.AIglm-5-2
2Ornith-1.0-397BDeepReinforce AIornith-1-0-397b
3Qwen3.7 MaxAlibabaqwen3-7-max
4Claude Opus 4.5Anthropic · Closed
5Qwen 3.6 Max (preview)Alibaba · Closed
+8 morereact Native Evals
16 models
1Composer 2Cursorcomposer-2
2Composer 2 FastCursorcomposer-2-fast
3GPT-5.4OpenAIgpt-5-4
4GPT-5.5OpenAI · Closed
5Claude Opus 4.6Anthropic · Closed
+11 moreswe Verified Arcee
5 models
1Claude Opus 4.6Anthropicclaude-opus-4-6
2MiniMax M2.7MiniMaxminimax-m2-7
3GLM-5Z.AIglm-5
4Kimi K2.5Moonshot AI · Open weight
5Trinity-Large-ThinkingArcee AI · Open weight
sci Code
13 models
1Sakana FuguSakana AIsakana-fugu
2Sakana Fugu-UltraSakana AIsakana-fugu-ultra
3Qwen3.7 MaxAlibabaqwen3-7-max
4Gemini 3.5 FlashGoogle · Closed
5Kimi K2.6Moonshot AI · Open weight
+8 moreaa Coding Index
92 models
1Claude Opus 5Anthropicclaude-opus-5
2GPT-5.6 SolOpenAIgpt-5-6-sol
3GPT-5.6 TerraOpenAIgpt-5-6-terra
4Claude Fable 5Anthropic · Closed
5Kimi K3Moonshot AI · Closed
+87 moreaa Sci Code
163 models
1Claude Fable 5Anthropicclaude-fable-5
2Gemini 3.1 ProGooglegemini-3-1-pro
3Kimi K3Moonshot AIkimi-3
4Muse Spark 1.1Meta · Closed
5GPT-5.4OpenAI · Closed
+158 moreAgentic38 benchmarks
terminal Bench2
45 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2Claude Mythos 5Anthropicclaude-mythos-5
3GPT-5.6 TerraOpenAIgpt-5-6-terra
4GPT-5.6 LunaOpenAI · Closed
5Claude Fable 5Anthropic · Closed
+40 morebrowse Comp
33 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2Kimi K3Moonshot AIkimi-3
3Claude Opus 5Anthropicclaude-opus-5
4GPT-5.5 ProOpenAI · Closed
5GPT-5.4 ProOpenAI · Closed
+28 morehle With Tools
10 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Sonnet 5Anthropicclaude-sonnet-5
3Qwen3.7 MaxAlibabaqwen3-7-max
4DeepSeek V4 Pro (Max)DeepSeek · Open weight
5Agents-A1InternScience · Open weight
+5 moregdpval Aa
87 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Fable 5Anthropicclaude-fable-5
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Kimi K3Moonshot AI · Closed
5Claude Sonnet 5Anthropic · Closed
+82 moregdpval Aa Normalized
87 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Fable 5Anthropicclaude-fable-5
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Kimi K3Moonshot AI · Closed
5Claude Sonnet 5Anthropic · Closed
+82 moreaa Agentic Index
86 models
1Claude Opus 5Anthropicclaude-opus-5
2GPT-5.6 SolOpenAIgpt-5-6-sol
3Claude Fable 5Anthropicclaude-fable-5
4Kimi K3Moonshot AI · Closed
5GPT-5.6 TerraOpenAI · Closed
+81 moreapex Agents Aa
27 models
1Gemini 3.5 FlashGooglegemini-3-5-flash
2Kimi K3Moonshot AIkimi-3
3GPT-5.6 TerraOpenAIgpt-5-6-terra
4GPT-5.5OpenAI · Closed
5GPT-5.6 LunaOpenAI · Closed
+22 moregert Labs
56 models
1Claude Opus 4.8Anthropicclaude-opus-4-8
2GPT-5.5OpenAIgpt-5-5
3Claude Opus 4.7Anthropicclaude-opus-4-7
4GPT-5.4OpenAI · Closed
5Qwen3.7 MaxAlibaba · Closed
+51 moreos World Verified
27 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Claude Fable 5Anthropicclaude-fable-5
3Claude Opus 4.8Anthropicclaude-opus-4-8
4Gemini 3.6 FlashGoogle · Closed
5Holo3-35B-A3BH Company · Open weight
+22 morecyber Gym
14 models
1Fugu CyberSakana AIsakana-fugu-cyber
2GPT-5.6 SolOpenAIgpt-5-6-sol
4GPT-5.6 TerraOpenAI · Closed
5GPT-5.5OpenAI · Closed
6GPT-5.4OpenAI · Closed
+9 moreandroid World
5 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Holo3.1-35B-A3BH Companyholo3-1-35b-a3b
3Holo3.1-4BH Companyholo3-1-4b
4Holo3.1-9BH Company · Open weight
5Qwen3.6-27BAlibaba · Open weight
mcp Atlas
30 models
1Muse Spark 1.1Metamuse-spark-1-1
2Claude Opus 5Anthropicclaude-opus-5
3Kimi K3Moonshot AIkimi-3
4Gemini 3.5 FlashGoogle · Closed
5Claude Opus 4.8Anthropic · Closed
+25 moretoolathlon
26 models
1Muse Spark 1.1Metamuse-spark-1-1
2Claude Opus 4.8Anthropicclaude-opus-4-8
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Gemini 3.5 FlashGoogle · Closed
5GPT-5.5OpenAI · Closed
+21 moretau2 Bench
142 models
4Step 3.7 FlashStepFun · Open weight98.5%
5GLM-5-TurboZ.AI · Closed98.5%
6GLM-5V-TurboZ.AI · Closed98.5%
7GLM-5Z.AI · Open weight98.2%
8GLM-5.1Z.AI · Open weight97.7%
+137 moredeep Search Qa
12 models
1Claude Opus 5Anthropicclaude-opus-5
2Kimi K3Moonshot AIkimi-3
3Claude Opus 4.8Anthropicclaude-opus-4-8
4Step 3.7 FlashStepFun · Open weight
5Kimi K2.6Moonshot AI · Open weight
+7 moretau2 Airline
0 models
pinch Bench
43 models
1Anthropic93.5%Overall
2Qwen3.7 MaxAlibabaqwen/qwen3.7-max
5MiMo-V2.5Xiaomi · Closed
6Grok Build 0.1xAI · Closed
7GPT-5.6 LunaOpenAI · Closed
+38 moreopen Hands Index
33 models
1Claude Fable 5Anthropicclaude-fable-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3Claude Opus 4.7 (Adaptive)Anthropicclaude-opus-4-7
4Claude Opus 4.6Anthropic · Closed
5GPT-5.5OpenAI · Closed
+28 moreswe Atlas Refactoring
12 models
1Anthropic54.8%Overall
4GPT-5.5OpenAI · Closed
5GPT-5.4OpenAI · Closed
6GLM-5.2Z.AI · Open weight
7GPT-5.3 CodexOpenAI · Closed
+7 moreinference Bench
16 models
1Claude Opus 4.8 (xHigh)Anthropic7.34x
2GPT-5.4 (High)OpenAI6.16x
3Claude Sonnet 4.6Anthropic5.56x
4GPT-5.3 Codex (High)OpenAI · Closed
5GPT-5.5 (xHigh)OpenAI · Closed
+11 morebfcl V4
9 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Qwen3.7 PlusAlibabaqwen3-7-plus
3LFM2.5-8B-A1BLiquidAIlfm2-5-8b-a1b
4Mellum2-12B-A2.5B-ThinkingJetBrains · Open weight
5Mellum2-12B-A2.5B-InstructJetBrains · Open weight
+4 moreclaw Eval
26 models
1Claude Opus 4.6Anthropicopus46
2Step 3.7 FlashStepFunstep_37_flash
3Claude Sonnet 4.6Anthropicsonnet46
4MiMo-V2.5-ProXiaomi · Closed
5Muse SparkMeta · Closed
+21 moreqwen Claw Bench
10 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen 3.6 Max (preview)Alibabaqwen3-6-max-preview
4Qwen3.6 PlusAlibaba · Closed
5Kimi K2.5Moonshot AI · Open weight
+5 moreqwen Web Bench
5 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen 3.6 Max (preview)Alibabaqwen3-6-max-preview
4Qwen3.6-27BAlibaba · Open weight
5Qwen3.6-35B-A3BAlibaba · Open weight
tau3 Bench
8 models
1Mistral Medium 3.5 128BMistralτ³-bench Telecom
4Qwen3.6 PlusAlibaba · Closed70.7%
5GLM-5.1Z.AI · Open weight70.6%
6Claude Opus 4.5Anthropic · Closed70.2%
7Qwen3.5 397BAlibaba · Open weight68.4%
+3 morevita Bench
10 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen3.6 PlusAlibabaqwen3-6-plus
4Qwen3.5 397BAlibaba · Open weight
5Agents-A1InternScience · Open weight
+5 moredeep Planning
7 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Qwen3.6 PlusAlibabaqwen3-6-plus
3Qwen3.5 397BAlibabaqwen3-5-397b
4Claude Opus 4.5Anthropic · Closed
5Qwen3.6-35B-A3BAlibaba · Open weight
+2 moremcp Tasks
5 models
1Qwen3.5 397BAlibabaqwen3-5-397b
2Qwen3.6 PlusAlibabaqwen3-6-plus
3Claude Opus 4.5Anthropicclaude-opus-4-5
4GLM-5Z.AI · Open weight
5Kimi K2.5Moonshot AI · Open weight
wide Research
7 models
1Kimi K2.6Moonshot AIkimi-2-6
2Claude Opus 4.5Anthropicclaude-opus-4-5
3Qwen3.6 PlusAlibabaqwen3-6-plus
4Qwen3.5 397BAlibaba · Open weight
5Kimi K2.5Moonshot AI · Open weight
+2 morefinance Agent V2
40 models
1Anthropicanthropic/claude-opus-558.6%
2Googlegoogle/gemini-3.5-flash57.9%
3Metameta/muse_spark_1_157.2%
4Claude Fable 5Anthropic
5Gemini 3.6 FlashGoogle
+35 moreReasoning22 benchmarks
bbh
4 models
1DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
3Soofi S 30B-A3BSoofi Projectsoofi-s-30b-a3b
4MiniCPM5-1BOpenBMB · Open weight
5Gemma 4 12BGoogle · Open weight
drop
2 models
1DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
3Soofi S 30B-A3BSoofi Projectsoofi-s-30b-a3b
lisan Bench
143 models
3Anthropicanthropic/claude-opus-4.6:thinking-16k3526.49
4GPT 5.5 (medium)OpenAI · Closed
5Sonnet 4.6 (16k)Anthropic
6GPT 5.4 (medium)OpenAI · Closed
7Opus 4.8 (high)Anthropic · Closed
+138 morepp Bench
63 models
1Claude Fable 5Anthropicclaude-fable-5@high
2GPT-5.5OpenAIgpt-5.5@xhigh
3GPT-5.4OpenAIgpt-5.4@xhigh
4GPT-5.2OpenAI · Closed
5Claude Opus 4.7Anthropic · Closed
+58 morelong Bench V2
12 models
1Claude Opus 4.5Anthropicclaude-opus-4-5
2Qwen3.5 397BAlibabaqwen3-5-397b
3Qwen3.6 PlusAlibabaqwen3-6-plus
4Nemotron 3 UltraNVIDIA · Open weight
5Kimi K2.5Moonshot AI · Open weight
+7 moremrcrv2
7 models
1Sakana Fugu-UltraSakana AIsakana-fugu-ultra
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen3.7 MaxAlibabaqwen3-7-max
4Sakana FuguSakana AI · Closed
5Gemini 3.5 FlashGoogle · Closed
+2 moremrcr1m
7 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
2DeepSeek V4 Pro (High)DeepSeekdeepseek-v4-pro-high
4DeepSeek V4 Flash (High)DeepSeek · Open weight
5Muse Spark 1.1Meta · Closed
6DeepSeek V4 ProDeepSeek · Open weight
+2 morecorpus Qa1m
4 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
4DeepSeek V4 Pro (High)DeepSeek · Open weight
5DeepSeek V4 ProDeepSeek · Open weight
6DeepSeek V4 FlashDeepSeek · Open weight
arc Agi2
19 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2Claude Opus 5Anthropicclaude-opus-5
3GPT-5.5OpenAIgpt-5-5
4GPT-5.6 TerraOpenAI · Closed
5GPT-5.4 ProOpenAI · Closed
+14 moreai Needle
4 models
1Claude Opus 4.5Anthropicclaude-opus-4-5
2Qwen3.5 397BAlibabaqwen3-5-397b
3Qwen3.6 PlusAlibabaqwen3-6-plus
4GLM-5Z.AI · Open weight
gpqa Diamond
45 models
1Sakana Fugu-UltraSakana AIsakana-fugu-ultra
2Sakana FuguSakana AIsakana-fugu
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Gemini 3.1 ProGoogle · Closed
5Claude Opus 4.7 (Adaptive)Anthropic · Closed
+40 morelcr
156 models
1GPT-5.2-CodexOpenAIgpt-5-2-codex
2GPT-5 (high)OpenAIgpt-5-high
3GPT-5.1OpenAIgpt-5-1
4Kimi K3Moonshot AI · Closed
5GPT-5.5OpenAI · Closed
+151 morecritpt
158 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2GPT-5.5 ProOpenAIgpt-5-5-pro
3GPT-5.6 TerraOpenAIgpt-5-6-terra
4GPT-5.4 ProOpenAI · Closed
5Claude Opus 5Anthropic · Closed
+153 morebullshit Bench V2
77 models
4Claude Opus 4.5 (high)Anthropic · Closed
6Claude Opus 4.6 (high)Anthropic · Closed
8Claude Opus 4.7 (none)Anthropic · Closed
9Claude Sonnet 5 (low)Anthropic · Closed
10Claude Sonnet 4.5 (high)Anthropic · Closed
+72 moreKnowledge30 benchmarks
mmlu
8 models
1o1OpenAI91.8%
2GPT-4.1OpenAIgpt-4-1
3DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
4DeepSeek V4 Flash BaseDeepSeek · Open weight
5GPT-4.1 miniOpenAI · Closed
+3 moregpqa
71 models
1Sakana Fugu-UltraSakana AIsakana-fugu-ultra
2Sakana FuguSakana AIsakana-fugu
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Claude Opus 4.7 (Adaptive)Anthropic · Closed
5Claude Mythos 5Anthropic · Closed
+66 moresuper Gpqa
19 models
1Claude Opus 4.6Anthropicclaude-opus-4-6
2Claude Sonnet 4.6Anthropicclaude-sonnet-4-6
3Qwen 3.6 Max (preview)Alibabaqwen3-6-max-preview
4Qwen3.7 MaxAlibaba · Closed
5Qwen3.6 PlusAlibaba · Closed
+14 moremmlu Pro
43 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Claude Opus 4.5Anthropicclaude-opus-4-5
3Qwen3.7 PlusAlibabaqwen3-7-plus
4Qwen3.6 PlusAlibaba · Closed
5Qwen3.5 397BAlibaba · Open weight
+38 moreagieval
2 models
1DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
3Soofi S 30B-A3BSoofi Projectsoofi-s-30b-a3b
hle
45 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Mythos 5Anthropicclaude-mythos-5
3Muse Spark 1.1Metamuse-spark-1-1
4GPT-5.4 ProOpenAI · Closed
5Claude Opus 4.8Anthropic · Closed
+40 moreartificial Analysis
167 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Fable 5Anthropicclaude-fable-5
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Kimi K3Moonshot AI · Closed
5Claude Opus 4.8Anthropic · Closed
+162 moreaa Gpqa Diamond
163 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2Gemini 3.1 ProGooglegemini-3-1-pro
3Kimi K3Moonshot AIkimi-3
4GPT-5.5OpenAI · Closed
5Claude Opus 5Anthropic · Closed
+158 moreaa Hle
163 models
1Claude Fable 5Anthropicclaude-fable-5
2Claude Opus 5Anthropicclaude-opus-5
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Claude Opus 4.8Anthropic · Closed
5Muse Spark 1.1Meta · Closed
+158 moreaa Omniscience Index
154 models
1Claude Fable 5Anthropicclaude-fable-5
2Gemini 3.1 ProGooglegemini-3-1-pro
3Claude Opus 5Anthropicclaude-opus-5
4Claude Opus 4.8Anthropic · Closed
5Grok 4.5xAI · Closed
+149 moreomniscience Accuracy
154 models
1Claude Fable 5Anthropicclaude-fable-5
2GPT-5.6 SolOpenAIgpt-5-6-sol
3GPT-5.5OpenAIgpt-5-5
4Gemini 3 ProGoogle · Closed
5Gemini 3.1 ProGoogle · Closed
+149 moreomniscience Hallucination Rate
154 models
1Command A+Coherecommand-a-plus
2MiniMax M3MiniMaxminimax-m3
3Qwen3.7 MaxAlibabaqwen3-7-max
4MiMo-V2.5-ProXiaomi · Closed
5Grok 4.3xAI · Closed
+149 moresimple Qa
9 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
2DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
3DeepSeek V4 Pro (High)DeepSeekdeepseek-v4-pro-high
4DeepSeek V4 ProDeepSeek · Open weight
5DeepSeek V4 Flash (Max)DeepSeek · Open weight
+4 morechinese Simple Qa
5 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
3DeepSeek V4 Pro (High)DeepSeekdeepseek-v4-pro-high
4DeepSeek V4 ProDeepSeek · Open weight
5DeepSeek V4 Flash (High)DeepSeek · Open weight
6DeepSeek V4 FlashDeepSeek · Open weight
health Bench Hard
8 models
1Muse SparkMetamuse-spark
2GPT-5.4OpenAIgpt-5-4
3GPT-5.6 SolOpenAIgpt-5-6-sol
4GPT-5.6 TerraOpenAI · Closed
5GPT-5.6 LunaOpenAI · Closed
+3 moremed Xpert Qa Text
5 models
1Gemini 3.1 ProGooglegemini-3-1-pro
2GPT-5.4OpenAIgpt-5-4
3Muse SparkMetamuse-spark
4Claude Opus 4.6Anthropic · Closed
5Grok 4.20xAI · Closed
hle No Tools
26 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Claude Opus 5Anthropicclaude-opus-5
3Muse Spark 1.1Metamuse-spark-1-1
4Sakana Fugu-UltraSakana AI · Closed
5Claude Opus 4.8Anthropic · Closed
+21 moremmlu Pro Arcee
6 models
1Claude Opus 4.6Anthropicclaude-opus-4-6
2Kimi K2.5Moonshot AIkimi-k2-5
3GLM-5Z.AIglm-5
4Trinity-Large-ThinkingArcee AI · Open weight
5MiniMax M2.7MiniMax · Open weight
+1 moremmlu Redux
11 models
1Claude Opus 4.5Anthropicclaude-opus-4-5
2Qwen3.7 MaxAlibabaqwen3-7-max
3Qwen3.5 397BAlibabaqwen3-5-397b
4Qwen3.7 PlusAlibaba · Closed
5Qwen3.6 PlusAlibaba · Closed
+6 moremmmlu
6 models
1Interfaze BetaInterfazeinterfaze-beta
2Qwen3.7 MaxAlibabaqwen3-7-max
3DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
4Qwen3.7 PlusAlibaba · Closed
5DeepSeek V4 Flash BaseDeepSeek · Open weight
+1 moreMultimodal44 benchmarks
mmmu
9 models
1Qwen3.6 PlusAlibabaqwen3-6-plus
2Qwen3.5-122B-A10BAlibabaqwen3-5-122b-a10b
3Qwen3.6-27BAlibabaqwen3-6-27b
4Qwen3.5-27BAlibaba · Open weight
5Qwen3.6-35B-A3BAlibaba · Open weight
+4 moremmmu Pro
34 models
1GPT-5.4 ProOpenAIgpt-5-4-pro
2Gemini 3.1 ProGooglegemini-3-1-pro
3Gemini 3.5 FlashGooglegemini-3-5-flash
4GPT-5.6 SolOpenAI · Closed
5Kimi K3Moonshot AI · Closed
+29 moreaa Mmmu Pro
92 models
1Claude Opus 5Anthropicclaude-opus-5
2Gemini 3.5 FlashGooglegemini-3-5-flash
3GPT-5.6 SolOpenAIgpt-5-6-sol
4Gemini 3.6 FlashGoogle · Closed
5Gemini 3.1 ProGoogle · Closed
+87 moredesign Arena Website
87 models
1Kimi K3Moonshot AIkimi-3
2GLM-5.2Z.AIglm-5-2
3Claude Fable 5Anthropicclaude-fable-5
4Claude Opus 4.7 (Adaptive)Anthropic · Closed
5Claude Opus 4.6Anthropic · Closed
+82 moreoffice Qa Pro
8 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3Kimi K3Moonshot AIkimi-3
4Claude Fable 5Anthropic · Closed
5GPT-5.5OpenAI · Closed
+3 moremmmu Pro Python
9 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2Kimi K3Moonshot AIkimi-3
3GPT-5.5OpenAIgpt-5-5
4GPT-5.4OpenAI · Closed
5GPT-5.6 TerraOpenAI · Closed
+4 moreomni Doc Bench15
3 models
1MiniMax M3MiniMaxminimax-m3
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen3.6-35B-A3BAlibabaqwen3-6-35b-a3b
real World Qa
4 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Qwen3.6-35B-A3BAlibabaqwen3-6-35b-a3b
3Qwen3.6-27BAlibabaqwen3-6-27b
4LFM2.5-VL-450MLiquidAI · Open weight
video Mme With Sub
5 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Qwen3.6-27BAlibabaqwen3-6-27b
3MiMo-V2.5Xiaomimimo-v2-5
4Qwen3.6-35B-A3BAlibaba · Open weight
5MiniMax M3MiniMax · Open weight
math Vision
12 models
1Kimi K3Moonshot AIkimi-3
2Qwen3.7 PlusAlibabaqwen3-7-plus
3Qwen3.5 397BAlibabaqwen3-5-397b
4Qwen3.6 PlusAlibaba · Closed
5Kimi K2.6Moonshot AI · Open weight
+7 moremm Long Bench Doc
0 models
refcoco Avg
3 models
1Qwen3.6-27BAlibabaqwen3-6-27b
2Qwen3.6-35B-A3BAlibabaqwen3-6-35b-a3b
4Interfaze BetaInterfaze · Closed
erqa
7 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Gemini 3.1 ProGooglegemini-3-1-pro
3GPT-5.4OpenAIgpt-5-4
4Muse SparkMeta · Closed
5Qwen3.6-27BAlibaba · Open weight
+2 morevideo Mmmu
9 models
1Gemini 3 ProGooglegemini-3-pro
2Kimi K2.5Moonshot AIkimi-k2-5
3Qwen3.7 PlusAlibabaqwen3-7-plus
4Qwen3.5 397BAlibaba · Open weight
5MiniMax M3MiniMax · Open weight
+4 moremlvu Avg
3 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Qwen3.6-27BAlibabaqwen3-6-27b
3Qwen3.6-35B-A3BAlibabaqwen3-6-35b-a3b
mmvu
4 models
1Kimi K2.5Moonshot AIkimi-k2-5
2Qwen3.5-122B-A10BAlibabaqwen3-5-122b-a10b
3Qwen3.5-27BAlibabaqwen3-5-27b
4Qwen3.5-35B-A3BAlibaba · Open weight
screen Spot Pro
15 models
1Claude Opus 4.8Anthropicclaude-opus-4-8
2GPT-5.4OpenAIgpt-5-4
3Gemini 3.1 ProGooglegemini-3-1-pro
4Muse SparkMeta · Closed
5Claude Opus 4.6Anthropic · Closed
+10 moremed Xpert Qa Mm
7 models
1Gemini 3.1 ProGooglegemini-3-1-pro
2Muse SparkMetamuse-spark
3GPT-5.4OpenAIgpt-5-4
4Qwen3.7 PlusAlibaba · Closed
5Grok 4.20xAI · Closed
+2 morezero Bench
4 models
1GPT-5.4OpenAIgpt-5-4
2Muse SparkMetamuse-spark
3Gemini 3.1 ProGooglegemini-3-1-pro
4Kimi K3Moonshot AI · Closed
simple Vqa
8 models
1Qwen3.7 PlusAlibabaqwen3-7-plus
2Step 3.7 FlashStepFunstep-3-7-flash
3Gemini 3.1 ProGooglegemini-3-1-pro
4Muse SparkMeta · Closed
5GPT-5.4OpenAI · Closed
+3 morev Star
11 models
1Kimi K2.6Moonshot AIkimi-2-6
2Qwen3.6 PlusAlibabaqwen3-6-plus
3Qwen3.5 397BAlibabaqwen3-5-397b
4Step 3.7 FlashStepFun · Open weight
5Qwen3.6-27BAlibaba · Open weight
+6 morecharxiv
29 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Kimi K3Moonshot AIkimi-3
3Claude Opus 4.7 (Adaptive)Anthropicclaude-opus-4-7-max
4Claude Opus 4.8Anthropic · Closed
5Muse Spark 1.1Meta · Closed
+24 morecharxiv No Tools
6 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Kimi K3Moonshot AIkimi-3
3Claude Opus 4.7 (Adaptive)Anthropicclaude-opus-4-7-max
4Claude Opus 4.8Anthropic · Closed
5InklingThinking Machines Lab · Open weight
+1 moreblueprint Bench2
2 models
1Claude Fable 5Anthropicclaude-fable-5
2Gemini 3.5 FlashGooglegemini-3-5-flash
Math23 benchmarks
aime2025
11 models
1MAI-Thinking-1Microsoftmai-thinking-1
2Kimi K2.5Moonshot AIkimi-k2-5
3Kimi K2.5 (Reasoning)Moonshot AIkimi-k2-5-reasoning
4GLM-4.7Z.AI · Open weight
5MiMo-V2-FlashXiaomi · Open weight
+6 moregsm8k
2 models
1DeepSeek V4 Pro BaseDeepSeekdeepseek-v4-pro-base
3Soofi S 30B-A3BSoofi Projectsoofi-s-30b-a3b
aime2025 Arcee
5 models
1Claude Opus 4.6Anthropicclaude-opus-4-6
2Kimi K2.5Moonshot AIkimi-k2-5
4GLM-5Z.AI · Open weight
5MiniMax M2.7MiniMax · Open weight
6Trinity-Large-PreviewArcee AI · Open weight
aime2026
17 models
1GLM-5.2Z.AIglm-5-2
2InklingThinking Machines Labinkling
3Kimi K2.6Moonshot AIkimi-2-6
4GLM-5Z.AI · Open weight
5Kimi K2.5Moonshot AI · Open weight
+12 morehmmt Feb2025
7 models
1GLM-5Z.AIglm-5
2Qwen3.6 PlusAlibabaqwen3-6-plus
3Kimi K2.5Moonshot AIkimi-k2-5
4Qwen3.5 397BAlibaba · Open weight
5Qwen3.6-27BAlibaba · Open weight
+2 morehmmt Nov2025
9 models
1GLM-5Z.AIglm-5
2Qwen3.6 PlusAlibabaqwen3-6-plus
3GLM-5.2Z.AIglm-5-2
4GLM-5.1Z.AI · Open weight
5Claude Opus 4.5Anthropic · Closed
+4 morehmmt Feb2026
20 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
4DeepSeek V4 Pro (High)DeepSeek · Open weight
5Qwen3.7 PlusAlibaba · Closed
6Kimi K2.6Moonshot AI · Open weight
+15 moreimo Answer Bench
8 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
4DeepSeek V4 Pro (High)DeepSeek · Open weight
5Qwen3.7 PlusAlibaba · Closed
6DeepSeek V4 Flash (High)DeepSeek · Open weight
+3 moreapex
8 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
4ZAYA1-8BZyphra · Open weight
5DeepSeek V4 Pro (High)DeepSeek · Open weight
6Qwen3.7 PlusAlibaba · Closed
+3 moreapex Shortlist
5 models
1DeepSeek V4 Pro (Max)DeepSeekdeepseek-v4-pro-max
3DeepSeek V4 Pro (High)DeepSeekdeepseek-v4-pro-high
4DeepSeek V4 Flash (High)DeepSeek · Open weight
5DeepSeek V4 FlashDeepSeek · Open weight
6DeepSeek V4 ProDeepSeek · Open weight
mm Answer Bench
10 models
1GLM-5.2Z.AIglm-5-2
2Kimi K2.6Moonshot AIkimi-2-6
3Claude Opus 4.5Anthropicclaude-opus-4-5
4GLM-5.1Z.AI · Open weight
5Qwen3.6 PlusAlibaba · Closed
+5 morefrontier Math
7 models
1GPT-5.6 SolOpenAIgpt-5-6-sol
2GPT-5.6 TerraOpenAIgpt-5-6-terra
3GPT-5.6 LunaOpenAIgpt-5-6-luna
4GPT-5.5 ProOpenAI · Closed
5GPT-5.5OpenAI · Closed
+2 moreusamo2026
3 models
1Claude Mythos 5Anthropicclaude-mythos-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3MiniMax M3MiniMaxminimax-m3
Multilingual7 benchmarks
mmlu Pro X
12 models
1Qwen3.7 MaxAlibabaqwen3-7-max
2Claude Opus 4.5Anthropicclaude-opus-4-5
3Qwen3.7 PlusAlibabaqwen3-7-plus
4Qwen3.6 PlusAlibaba · Closed
5Qwen3.5 397BAlibaba · Open weight
+7 morenova63
7 models
1Qwen3.5 397BAlibabaqwen3-5-397b
2Qwen3.7 MaxAlibabaqwen3-7-max
3Qwen3.7 PlusAlibabaqwen3-7-plus
4Qwen3.6 PlusAlibaba · Closed
5Claude Opus 4.5Anthropic · Closed
+2 moreinclude
4 models
1Claude Opus 5Anthropicclaude-opus-5
2Claude Opus 4.8Anthropicclaude-opus-4-8
3Qwen3.7 MaxAlibabaqwen3-7-max
4Qwen3.7 PlusAlibaba · Closed
Instruction Following4 benchmarks
ifeval
21 models
4Qwen3.7 MaxAlibaba · Closed94.3%
5Qwen3.6 PlusAlibaba · Closed94.3%
6Kimi K2.5Moonshot AI · Open weight93.9%
7o3-miniOpenAI · Closed93.9%
8Qwen3.5-122B-A10BAlibaba · Open weight93.4%
+16 moreif Bench
15 models
1MAI-Thinking-1Microsoftmai-thinking-1
3Grok 4.3xAIgrok-4-3
4InklingThinking Machines Lab · Open weight
5Qwen3.7 MaxAlibaba · Closed
6Qwen3.7 PlusAlibaba · Closed
+10 moreaa If Bench
145 models
1MiniMax M3MiniMaxminimax-m3
3Grok 4.3xAIgrok-4-3
4Qwen3.7 MaxAlibaba · Closed
5MiMo-V2.5-ProXiaomi · Closed
6DeepSeek V4 Flash (Max)DeepSeek · Open weight
+140 moreExternal15 benchmarks
vals Index
40 models
1Anthropicanthropic/claude-fable-575.14%
2Anthropicanthropic/claude-opus-574.82%
3Moonshot AIkimi/kimi-k374.70%
4GPT-5.6 SolOpenAI
5Claude Opus 4.8Anthropic
+35 morevals Multimodal Index
29 models
1Anthropicanthropic/claude-fable-574.15%
2Anthropicanthropic/claude-opus-573.90%
3Moonshot AIkimi/kimi-k373.42%
4GPT-5.6 SolOpenAI
5Claude Opus 4.8Anthropic
+24 morevals Corp Fin V2
126 models
1Anthropicanthropic/claude-opus-573.19%
2Anthropicanthropic/claude-fable-571.83%
3Moonshot AIkimi/kimi-k371.56%
4Muse Spark 1.1Meta
5InklingThinkingmachines
+121 morevals Med Code
76 models
1Anthropicanthropic/claude-opus-563.57%
2Googlegoogle/gemini-3.1-pro-preview59.06%
3Anthropicanthropic/claude-fable-556.07%
4Gemini 3 Flash PreviewGoogle
5Gemini 3.5 FlashGoogle
+71 morevals Med Scribe
75 models
1Anthropicanthropic/claude-opus-590.98%
2Metameta/muse_spark_1_188.89%
3Anthropicanthropic/claude-fable-588.52%
4GPT-5.1OpenAI
5Kimi K3Moonshot AI
+70 morevals Mortgage Tax
90 models
1Anthropicanthropic/claude-opus-572.06%
2Anthropicanthropic/claude-opus-4-770.27%
3Anthropicanthropic/claude-sonnet-570.03%
4Claude Opus 4.8Anthropic
5Gemini 3.1 Pro PreviewGoogle
+85 morevals Proof Bench
54 models
1Anthropicanthropic/claude-opus-578.00%
2OpenAIopenai/gpt-5.6-sol77.00%
3Anthropicanthropic/claude-fable-577.00%
4GPT-5.6 TerraOpenAI
5AristotleHarmonic
+49 morevals Legal Bench
129 models
1Anthropicanthropic/claude-fable-588.56%
2Googlegoogle/gemini-3.1-pro-preview87.40%
3Googlegoogle/gemini-3-pro-preview87.03%
4Claude Opus 5Anthropic
5GPT-5.6 SolOpenAI
+124 morevals Case Law V2
54 models
1xAIgrok/grok-4.379.31%
2OpenAIopenai/gpt-5.1-2025-11-1373.42%
3OpenAIopenai/gpt-4.1-2025-04-1469.88%
4GPT-5 MiniOpenAI
5Claude Opus 4.7Anthropic
+49 moredeep Swe
11 models
1OpenAImini_swe_agent_gpt_5_6_sol_max72.7%
4kimi-k3[max]Moonshot AI · Closed
6gpt-5-5[xhigh]OpenAI · Closed
7claude-opus-4-8[max]Anthropic · Closed
8claude-sonnet-5[max]Anthropic · Closed
+6 morevals Swe Bench
75 models
1Anthropicanthropic/claude-opus-597.00%
2OpenAIopenai/gpt-5.6-sol96.20%
3Anthropicanthropic/claude-fable-595.00%
4Kimi K3Moonshot AI
5GPT-5.6 LunaOpenAI
+70 morevals Terminal Bench2
67 models
1OpenAIopenai/gpt-5.573.20%
2Anthropicanthropic/claude-opus-4-870.04%
3Anthropicanthropic/claude-opus-4-768.54%
4Gemini 3.5 FlashGoogle
5Gemini 3.1 Pro PreviewGoogle
+62 morevals Live Code Bench
131 models
1Anthropicanthropic/claude-fable-589.78%
2Anthropicanthropic/claude-opus-589.03%
3Googlegoogle/gemini-3.1-pro-preview88.48%
4Gemini 3.6 FlashGoogle
5GPT-5.2 CodexOpenAI
+126 morevals Gpqa Diamond
126 models
1Googlegoogle/gemini-3.1-pro-preview95.45%
2OpenAIopenai/gpt-5.6-sol95.20%
3Googlegoogle/gemini-3.6-flash93.43%
4Claude Opus 5Anthropic
5Claude Fable 5Anthropic
+121 morevals Mmlu Pro
125 models
1Anthropicanthropic/claude-opus-591.59%
2Anthropicanthropic/claude-fable-591.50%
3Googlegoogle/gemini-3.1-pro-preview90.99%
4Gemini 3 Pro PreviewGoogle
5Claude Opus 4.7Anthropic
+120 more