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
Arena Hard Prompts
Arena Score on the hard-prompt subset — not a percentagehigher is betterProvenanceindependent leaderboard100 modelsscores Aug 2, 2026
1Claude Fable 5Anthropic1534.471231—Aug 2, 2026leaderboard2Claude Opus 4.6Anthropic1526.71861—Aug 2, 2026leaderboard3Claude Opus 5Anthropic1519.6774via High61—Aug 2, 2026leaderboard4Claude Opus 4.7Anthropic1516.817461—Aug 2, 2026leaderboard5Muse Spark 1.1Meta1510.618355—Aug 2, 2026leaderboard6Gemini 3.1 Pro Preview Custom ToolsGoogle1506.9543251—Aug 2, 2026leaderboard7Claude Opus 4.8Anthropic1506.027560—Aug 2, 2026leaderboard8Kimi K3Moonshot AI2.8T1505.5029via Max108~0.86Aug 2, 2026leaderboard9GPT-5.6 SolOpenAI1504.9292via xHigh100—Aug 2, 2026leaderboard10Claude Sonnet 4.6Anthropic1502.9878100—Aug 2, 2026leaderboard11Gemini 3.6 FlashGoogle1501.7937400—Aug 2, 2026leaderboard12Claude Opus 4.5Anthropic1497.583960—Aug 2, 2026leaderboard13GPT-5.2 ChatOpenAI1496.9454214—Aug 2, 2026leaderboard14GPT-5.5OpenAI1495.2811150—Aug 2, 2026leaderboard15Qwen3.7 MaxQwen1495.2038399—Aug 2, 2026leaderboard16MiMo-V2.5-ProXiaomi1T1494.02822147~2.3Aug 2, 2026leaderboard17Gemini 3.5 FlashGoogle1492.9637via Medium332—Aug 2, 2026leaderboard18GLM 5.1Z.AI754B1491.4741498~3.1Aug 2, 2026leaderboard19Grok 4.5xAI1491.217249—Aug 2, 2026leaderboard20GPT-5.6 TerraOpenAI1489.0639via xHigh496—Aug 2, 2026leaderboard21GLM 5.2Z.AI753B1488.0022via Max992~3.1Aug 2, 2026leaderboard22GPT-5.4OpenAI1487.6242198—Aug 2, 2026leaderboard23Grok 4.20xAI1486.2451594—Aug 2, 2026leaderboard24Kimi K2.6Moonshot AI1.1T1484.8312599~2.2Aug 2, 2026leaderboard25Grok 4.20 Multi-AgentxAI1484.1166594—Aug 2, 2026leaderboard26Claude Sonnet 4.5Anthropic1483.293199—Aug 2, 2026leaderboard27Qwen3.6 Max PreviewQwen1481.2656240—Aug 2, 2026leaderboard28DeepSeek V4 ProDeepSeek1.6T1480.91881702~1.5Aug 2, 2026leaderboard29Qwen3.7 PlusQwen1478.40011155—Aug 2, 2026leaderboard30GLM 5Z.AI754B1477.8748770~3.1Aug 2, 2026leaderboard31Claude Opus 4.1Anthropic1477.244220—Aug 2, 2026leaderboard32Gemini 3.5 Flash LiteGoogle1475.59891180—Aug 2, 2026leaderboard33Hy3Tencent299B1474.58912793~7.8Aug 2, 2026leaderboard34GPT-5.6 LunaOpenAI1474.4623via xHigh1117—Aug 2, 2026leaderboard35Gemma 4 31BGoogle32.7B1473.20614333~71Aug 2, 2026leaderboard36Kimi K2 ThinkingMoonshot AI1.1T1471.7405via Thinking589~2.2Aug 2, 2026leaderboard37GPT-5.3 ChatOpenAI1471.3879105—Aug 2, 2026leaderboard38GPT-5.4 MiniOpenAI1470.5989via High654—Aug 2, 2026leaderboard39Qwen3.6 PlusQwen1467.2709752—Aug 2, 2026leaderboard40MiniMax M3MiniMax427B1466.6281528~5.4Aug 2, 2026leaderboard41Qwen3.5 397B A17BQwen403B1463.2519625~5.8Aug 2, 2026leaderboard42GLM 4.7Z.AI358B1462.8175836~6.5Aug 2, 2026leaderboard43Gemma 4 26B A4BGoogle26.5B1462.65464876~88Aug 2, 2026leaderboard44InklingThinking Machines952B1461.9665361~2.4Aug 2, 2026leaderboard45MiMo-V2.5Xiaomi311B1461.84766526~7.5Aug 2, 2026leaderboard46Kimi K2.5Moonshot AI1.1T1461.284731~2.2Aug 2, 2026leaderboard47DeepSeek V4 FlashDeepSeek291B1459.26788272~8Aug 2, 2026leaderboard48Qwen3 MaxQwen1459.1249374—Aug 2, 2026leaderboard49Gemini 2.5 ProGoogle1458.226292—Aug 2, 2026leaderboard50Grok 4.3xAI1456.7785583—Aug 2, 2026leaderboard51GPT-5.1OpenAI1455.3406291—Aug 2, 2026leaderboard52GLM 5V TurboZ.AI1452.6275363—Aug 2, 2026leaderboard53Qwen3 235B A22B Instruct 2507Qwen235B1447.84682632~9.8Aug 2, 2026leaderboard54DeepSeek V3.2 ExpDeepSeek685B1447.35543530~3.3Aug 2, 2026leaderboard55DeepSeek V3.2DeepSeek685B1447.15844656~3.3Aug 2, 2026leaderboard56Mistral Medium 3.5Mistral AI128B1445.692193~18Aug 2, 2026leaderboard57Gemini 3.1 Flash Lite PreviewGoogle1444.65661926—Aug 2, 2026leaderboard58MiniMax M2.7MiniMax229B1442.19511502~10Aug 2, 2026leaderboard59GLM 4.6Z.AI357B1441.972824~6.4Aug 2, 2026leaderboard60o3OpenAI1440.0905180—Aug 2, 2026leaderboard61Qwen3 VL 235B A22B InstructQwen236B1440.0131636~9.7Aug 2, 2026leaderboard62Claude Haiku 4.5Anthropic1439.6061288—Aug 2, 2026leaderboard63Hy3 previewTencent299B1437.95096847~7.6Aug 2, 2026leaderboard64Claude Opus 4Anthropic1435.093619—Aug 2, 2026leaderboard65R1 0528DeepSeek685B1433.4307667~3.3Aug 2, 2026leaderboard66Mistral Large 3 2512Mistral AI1433.072955—Aug 2, 2026leaderboard67GLM 4.5Z.AI358B1432.6225651~6.3Aug 2, 2026leaderboard68Qwen3.5-122B-A10BQwen125B1432.5929689~18Aug 2, 2026leaderboard69GPT-4.1OpenAI1430.8196179—Aug 2, 2026leaderboard70Qwen3.5-27BQwen27.8B1426.7187915~81Aug 2, 2026leaderboard71DeepSeek V3.1 TerminusDeepSeek685B1423.52851498~3.3Aug 2, 2026leaderboard72GPT-5.4 NanoOpenAI1422.0078via High2275—Aug 2, 2026leaderboard73Qwen3 Next 80B A3B InstructQwen81.3B1420.74621821~28Aug 2, 2026leaderboard74Gemini 2.5 FlashGoogle1419.65591136—Aug 2, 2026leaderboard75Qwen3 VL 235B A22B ThinkingQwen236B1418.3614via Thinking546~9.5Aug 2, 2026leaderboard76o1OpenAI1417.748824—Aug 2, 2026leaderboard77Qwen3 235B A22B Thinking 2507Qwen235B1417.6632948~9.6Aug 2, 2026leaderboard78Qwen3.5-FlashQwen1415.95845446—Aug 2, 2026leaderboard79MiniMax M2.5MiniMax229B1415.01441572~9.8Aug 2, 2026leaderboard80Qwen3 Coder 480B A35BQwen480B1414.05541414~4.7Aug 2, 2026leaderboard81Step 3.5 FlashStepFun199B1413.3484711~11Aug 2, 2026leaderboard82Qwen3.5-35B-A3BQwen36B1413.33341413~62Aug 2, 2026leaderboard83DeepSeek V3 0324DeepSeek685B1408.57431565~3.3Aug 2, 2026leaderboard84Qwen3 30B A3B Instruct 2507Qwen30.5B1407.06747289~73Aug 2, 2026leaderboard85o4 MiniOpenAI1405.6208319—Aug 2, 2026leaderboard86GPT-5 MiniOpenAI1403.0584via High1403—Aug 2, 2026leaderboard87GPT-4.1 MiniOpenAI1402.511877—Aug 2, 2026leaderboard88o3 Mini HighOpenAI1401.2872via High318—Aug 2, 2026leaderboard89Claude Sonnet 5Anthropic1398.2943140—Aug 2, 2026leaderboard90GLM 4.5 AirZ.AI111B1390.5241636~20Aug 2, 2026leaderboard91GLM 4.7 FlashZ.AI31.2B1386.7916346773Aug 2, 2026leaderboard92Trinity Large ThinkingArcee AI399B1385.6948via Thinking1732~5.5Aug 2, 2026leaderboard93Qwen3 Next 80B A3B ThinkingQwen81.3B1384.1209via Thinking1775~27Aug 2, 2026leaderboard94GLM 4.6VZ.AI108B1382.56291536~20Aug 2, 2026leaderboard95MiniMax M1MiniMax1380.9256628—Aug 2, 2026leaderboard96Nemotron 3 SuperNVIDIA124B1380.15863450~18Aug 2, 2026leaderboard97GLM 4.5VZ.AI108B1375.6516764~20Aug 2, 2026leaderboard98o3 MiniOpenAI1370.3136311—Aug 2, 2026leaderboard99MiniMax M2MiniMax229B1368.51151342~9.5Aug 2, 2026leaderboard100Qwen3 32BQwen32.8B1367.41124884~66Aug 2, 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.