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 · Tool use
Inverse-propensity score for calling the right tool and not inventing one — higher is better, as on the parent boardProvenance55 modelsscores Aug 11, 2026 – Sep 25, 2026
1Gemini 3.6 FlashGoogleScore0.011ScoredAug 11, 20262Kimi K2.6Moonshot AI1.1TScore0.008ScoredSep 5, 20263GPT-5.5OpenAIScore0.008ScoredSep 5, 20264Kimi K2.7 CodeMoonshot AI1.1TScore0.008ScoredSep 5, 20265Claude Opus 4.6AnthropicScore0.008ScoredSep 5, 20266Claude Opus 4.7AnthropicScore0.007ScoredSep 5, 20267Grok 4.3xAIScore0.007ScoredSep 5, 20268Grok Build 0.1xAIScore0.005ScoredSep 5, 20269GLM 5.3 FlashZ.ai321BScore0.004ScoredSep 25, 202610GLM 5.3Z.aiScore0.004ScoredSep 25, 202611Grok 4.6xAIScore0.004ScoredSep 25, 202612GPT-5.4OpenAIScore0.004ScoredSep 25, 202613Claude Fable 5AnthropicScore0.004ScoredSep 25, 202614GPT-5.6 LunaOpenAIScore0.004ScoredSep 25, 202615GPT-5.6 SolOpenAIScore0.004ScoredSep 25, 202616GPT-5.6 TerraOpenAIScore0.004ScoredSep 25, 202617Kimi K3Moonshot AI2.8TScore0.004ScoredSep 25, 202618GPT-6 SolOpenAIScore0.004ScoredSep 25, 202619Grok 4.5xAIScore0.004ScoredSep 25, 202620Grok 4.7xAIScore0.004ScoredSep 25, 202621GLM 5.2Z.ai753BScore0.004ScoredSep 25, 202622GPT-6 AstraOpenAIScore0.004ScoredSep 25, 202623Muse Spark 1.3MetaScore0.004ScoredSep 25, 202624Claude Fable 5.1AnthropicScore0.004ScoredSep 25, 202625Claude Opus 5AnthropicScore0.003ScoredSep 25, 202626DeepSeek V4 FlashDeepSeek291BScore0.003ScoredSep 15, 202627Muse Spark 1.1MetaScore0.003ScoredSep 25, 202628Muse Spark 1.2MetaScore0.003ScoredSep 25, 202629Claude Sonnet 4.6AnthropicScore0.003ScoredSep 15, 202630Nemotron 3 UltraNVIDIA561BScore0.002ScoredSep 5, 202631DeepSeek V4.1 FlashDeepSeek763BScore0.002ScoredSep 25, 202632Gemini 3.8 FlashGoogleScore0.002ScoredSep 25, 202633Gemini 3.5 FlashGoogleScore0.002ScoredSep 5, 202634MiniMax M2.7MiniMax229BScore0.002ScoredSep 15, 202635Claude Sonnet 5AnthropicScore0.002ScoredSep 25, 202636Hy4 previewTencent780BScore0ScoredSep 25, 202637Gemini 3.7 FlashGoogleScore0ScoredSep 25, 202638InklingThinking Machines952BScore0ScoredSep 25, 202639MiniMax M3MiniMax427BScore−0.001ScoredSep 25, 202640DeepSeek V4 ProDeepSeek1.6TScore−0.002ScoredSep 25, 202641Qwen3.7 MaxQwenScore−0.002ScoredSep 25, 202642Qwen3.8 27BQwen27.8BScore−0.003ScoredSep 25, 202643Gemini 3.1 Pro Preview Custom ToolsGoogleScore−0.004ScoredSep 25, 202644Inkling SmallThinking Machines266BScore−0.004ScoredSep 25, 202645Qwen3.8 FlashQwenScore−0.009ScoredSep 25, 202646MiMo-V2.5-ProXiaomi1TScore−0.009ScoredSep 25, 202647Qwen3.7 PlusQwenScore−0.01ScoredSep 25, 202648Qwen3.8 MaxQwenScore−0.011ScoredSep 25, 202649Gemini 3.5 Flash LiteGoogleScore−0.015ScoredSep 15, 202650GLM 5.1Z.ai754BScore−0.015ScoredSep 5, 202651Solar Pro 4UpstageScore−0.019ScoredSep 25, 202652Mistral Medium 3.5Mistral AI128BScore−0.023ScoredSep 25, 202653Hy3Tencent299BScore−0.029ScoredSep 25, 202654Claude Opus 4.8AnthropicScore−0.272ScoredSep 5, 202655Gemma 4 31BGoogle31.3BScore−0.284ScoredSep 5, 2026
