Leaderboards
Benchmarks
Every score carries its origin.
Arena Agent55Arena Agent · Recovery55Arena Agent · Steerability55Arena Agent · Task outcome55Arena Agent · Tool use55LiveBench Agentic Coding58SWE-bench Verified42
Arena Agent · Steerability
Inverse-propensity score for doing what it was asked and changing course when told — higher is betterProvenance55 modelsscores Aug 11, 2026 – Sep 25, 2026
1GPT-6 SolOpenAIScore0.157ScoredSep 25, 20262Claude Fable 5AnthropicScore0.116ScoredSep 25, 20263Claude Opus 5AnthropicScore0.112ScoredSep 25, 20264Claude Opus 4.8AnthropicScore0.094ScoredSep 5, 20265Kimi K2.6Moonshot AI1.1TScore0.077ScoredSep 5, 20266Claude Fable 5.1AnthropicScore0.074ScoredSep 25, 20267Claude Opus 4.7AnthropicScore0.073ScoredSep 5, 20268Claude Sonnet 5AnthropicScore0.072ScoredSep 25, 20269GPT-5.6 SolOpenAIScore0.066ScoredSep 25, 202610Kimi K2.7 CodeMoonshot AI1.1TScore0.047ScoredSep 5, 202611Claude Opus 4.6AnthropicScore0.046ScoredSep 5, 202612GLM 5.2Z.ai753BScore0.045ScoredSep 25, 202613GPT-5.6 TerraOpenAIScore0.043ScoredSep 25, 202614Grok 4.5xAIScore0.043ScoredSep 25, 202615Gemini 3.8 FlashGoogleScore0.037ScoredSep 25, 202616Grok 4.6xAIScore0.035ScoredSep 25, 202617GLM 5.3Z.aiScore0.03ScoredSep 25, 202618Qwen3.8 MaxQwenScore0.027ScoredSep 25, 202619GPT-5.5OpenAIScore0.027ScoredSep 25, 202620GPT-5.4OpenAIScore0.026ScoredSep 25, 202621GPT-5.6 LunaOpenAIScore0.018ScoredSep 25, 202622Muse Spark 1.3MetaScore0.017ScoredSep 25, 202623DeepSeek V4 FlashDeepSeek291BScore0.004ScoredSep 15, 202624Gemini 3.7 FlashGoogleScore0.001ScoredSep 25, 202625Kimi K3Moonshot AI2.8TScore−0.002ScoredSep 25, 202626Hy4 previewTencent780BScore−0.004ScoredSep 25, 202627GLM 5.3 FlashZ.ai321BScore−0.008ScoredSep 25, 202628Qwen3.8 27BQwen27.8BScore−0.009ScoredSep 25, 202629GPT-6 AstraOpenAIScore−0.009ScoredSep 25, 202630GLM 5.1Z.ai754BScore−0.013ScoredSep 5, 202631Hy3Tencent299BScore−0.016ScoredSep 25, 202632Grok 4.7xAIScore−0.024ScoredSep 25, 202633DeepSeek V4 ProDeepSeek1.6TScore−0.025ScoredSep 25, 202634Qwen3.7 MaxQwenScore−0.031ScoredSep 25, 202635Gemini 3.5 FlashGoogleScore−0.034ScoredSep 5, 202636DeepSeek V4.1 FlashDeepSeek763BScore−0.035ScoredSep 25, 202637Qwen3.8 FlashQwenScore−0.037ScoredSep 25, 202638Gemini 3.1 Pro Preview Custom ToolsGoogleScore−0.037ScoredSep 25, 202639MiMo-V2.5-ProXiaomi1TScore−0.043ScoredSep 25, 202640Gemini 3.6 FlashGoogleScore−0.049ScoredAug 11, 202641Claude Sonnet 4.6AnthropicScore−0.05ScoredSep 15, 202642Muse Spark 1.1MetaScore−0.053ScoredSep 25, 202643Muse Spark 1.2MetaScore−0.056ScoredSep 25, 202644MiniMax M2.7MiniMax229BScore−0.066ScoredSep 15, 202645Grok Build 0.1xAIScore−0.067ScoredSep 5, 202646Mistral Medium 3.5Mistral AI128BScore−0.068ScoredSep 25, 202647MiniMax M3MiniMax427BScore−0.07ScoredSep 25, 202648Qwen3.7 PlusQwenScore−0.07ScoredSep 25, 202649Nemotron 3 UltraNVIDIA561BScore−0.097ScoredSep 5, 202650Solar Pro 4UpstageScore−0.097ScoredSep 25, 202651Gemma 4 31BGoogle31.3BScore−0.107ScoredSep 5, 202652Grok 4.3xAIScore−0.111ScoredSep 5, 202653InklingThinking Machines952BScore−0.117ScoredSep 25, 202654Inkling SmallThinking Machines266BScore−0.119ScoredSep 25, 202655Gemini 3.5 Flash LiteGoogleScore−0.128ScoredSep 15, 2026