Leaderboards
Benchmarks
Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.
Arena Hard Prompts143GPQA Diamond16LiveBench35LiveBench Data Analysis35LiveBench Reasoning35MMLU0MMLU-Pro16
LiveBench Reasoning
% average over the reasoning tasks of the monthly-refreshed LiveBench sethigher is betterProvenanceindependent leaderboard35 modelsscores Jun 25, 2026
1GPT-5.6 SolOpenAI91.65via Max6—Jun 25, 2026leaderboard2Claude Opus 5Anthropic91.21via Max effort4—Jun 25, 2026leaderboard3Kimi K3Moonshot AI2.8T90.676~0.052Jun 25, 2026leaderboard4GPT-5.6 TerraOpenAI90.63via Max30—Jun 25, 2026leaderboard5GPT-5.5OpenAI89.65via xHigh9—Jun 25, 2026leaderboard6Claude Fable 5Anthropic89.65via Max effort2—Jun 25, 2026leaderboard7Claude Opus 4.8Anthropic89.19via Max effort4—Jun 25, 2026leaderboard8Claude Sonnet 5Anthropic88.69via xHigh effort9—Jun 25, 2026leaderboard9Claude Opus 4.6Anthropic88.67via Thinking Auto High effort4—Jun 25, 2026leaderboard10GPT-5.4OpenAI88.12via xHigh12—Jun 25, 2026leaderboard11Muse Spark 1.1Meta87.73via xHigh21—Jun 25, 2026leaderboard12Claude Opus 4.7Anthropic87.19via xHigh effort3—Jun 25, 2026leaderboard13Grok 4.5xAI87.1715—Jun 25, 2026leaderboard14GPT-5.6 LunaOpenAI85.64via Max65—Jun 25, 2026leaderboard15Gemini 3.6 FlashGoogle85.15via High23—Jun 25, 2026leaderboard16Claude Sonnet 4.6Anthropic84.77via Thinking Auto Medium effort6—Jun 25, 2026leaderboard17Qwen3.7 MaxQwen83.3422—Jun 25, 2026leaderboard18GPT-5.2 ChatOpenAI83.21via High12—Jun 25, 2026leaderboard19Kimi K2.7 CodeMoonshot AI1.1T82.8124~0.12Jun 25, 2026leaderboard20DeepSeek V4 ProDeepSeek1.6T82.6995~0.082Jun 25, 2026leaderboard21Gemini 3.5 FlashGoogle82via High18—Jun 25, 2026leaderboard22GPT-5.4 NanoOpenAI81.1via xHigh130—Jun 25, 2026leaderboard23Claude Opus 4.5Anthropic80.09via Thinking 64k High effort3—Jun 25, 2026leaderboard24Kimi K2.6Moonshot AI1.1T79.38via Thinking32~0.12Jun 25, 2026leaderboard25GLM 5.2Z.AI753B78.6352~0.17Jun 25, 2026leaderboard26InklingThinking Machines952B78.35via xHigh19~0.13Jun 25, 2026leaderboard27GPT-5.2-CodexOpenAI77.716—Jun 25, 2026leaderboard28Grok Build 0.1xAI76.3738—Jun 25, 2026leaderboard29Qwen3.6 PlusQwen75.8339—Jun 25, 2026leaderboard30MiniMax M3MiniMax427B74.4878~0.28Jun 25, 2026leaderboard31GPT-5.4 MiniOpenAI71.32via xHigh32—Jun 25, 2026leaderboard32Grok 4.3xAI70.8228—Jun 25, 2026leaderboard33DeepSeek V4 FlashDeepSeek291B70.58400~0.38Jun 25, 2026leaderboard34Qwen3.6 27BQwen27.8B70.2837~4Jun 25, 2026leaderboard35Gemini 3.5 Flash LiteGoogle60.19via High48—Jun 25, 2026leaderboard
A tilde in Score / GB VRAM means the memory figure behind it is calculated from the parameter count rather than read from a published file. Those ratios are shown to two significant figures, and rows within a few percent of each other are not distinguishable.