Leaderboards
Benchmarks
Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.
Arena Maths
Arena Score on maths prompts — not a percentagehigher is betterProvenanceindependent leaderboard100 modelsscores Aug 2, 2026
1Claude Opus 5Anthropic1552.5638via Max62—Aug 2, 2026leaderboard2Claude Fable 5Anthropic1534.422631—Aug 2, 2026leaderboard3Gemini 3.6 FlashGoogle1521.1993406—Aug 2, 2026leaderboard4Gemini 3.5 FlashGoogle1521.0412via High338—Aug 2, 2026leaderboard5Claude Opus 4.6Anthropic1504.499760—Aug 2, 2026leaderboard6GPT-5.5OpenAI1499.2765150—Aug 2, 2026leaderboard7Gemini 3.1 Pro Preview Custom ToolsGoogle1492.1811249—Aug 2, 2026leaderboard8Claude Opus 4.7Anthropic1491.368160—Aug 2, 2026leaderboard9Qwen3.7 MaxQwen1489.772397—Aug 2, 2026leaderboard10Muse Spark 1.1Meta1488.2358350—Aug 2, 2026leaderboard11Grok 4.5xAI1488.2356248—Aug 2, 2026leaderboard12GPT-5.6 SolOpenAI1485.3478via xHigh99—Aug 2, 2026leaderboard13GLM 5.1Z.AI754B1482.6234496~3.1Aug 2, 2026leaderboard14Kimi K2.6Moonshot AI1.1T1479.3165596~2.2Aug 2, 2026leaderboard15GPT-5.6 LunaOpenAI1475.975via xHigh1118—Aug 2, 2026leaderboard16GPT-5.6 TerraOpenAI1475.9289via xHigh492—Aug 2, 2026leaderboard17InklingThinking Machines952B1475.6546364~2.5Aug 2, 2026leaderboard18MiMo-V2.5-ProXiaomi1T1475.38372120~2.3Aug 2, 2026leaderboard19GLM 5.2Z.AI753B1475.0619via Max983~3.1Aug 2, 2026leaderboard20Qwen3.6 Max PreviewQwen1474.3032239—Aug 2, 2026leaderboard21Claude Opus 4.8Anthropic1472.472859—Aug 2, 2026leaderboard22Gemma 4 31BGoogle32.7B1471.30984327~71Aug 2, 2026leaderboard23Kimi K2 ThinkingMoonshot AI1.1T1469.8938via Thinking588~2.2Aug 2, 2026leaderboard24Qwen3.7 PlusQwen1469.76641148—Aug 2, 2026leaderboard25Gemma 4 26B A4BGoogle26.5B1467.54494892~88Aug 2, 2026leaderboard26Claude Opus 4.5Anthropic1464.980559—Aug 2, 2026leaderboard27Claude Sonnet 4.6Anthropic1463.806398—Aug 2, 2026leaderboard28GPT-5.4OpenAI1460.721195—Aug 2, 2026leaderboard29Qwen3.6 PlusQwen1455.1582746—Aug 2, 2026leaderboard30GPT-5.2 ChatOpenAI1453.7383208—Aug 2, 2026leaderboard31Grok 4.20 Multi-AgentxAI1453.0283581—Aug 2, 2026leaderboard32Grok 4.20xAI1452.0497581—Aug 2, 2026leaderboard33GLM 5V TurboZ.AI1448.6096362—Aug 2, 2026leaderboard34Qwen3.5 397B A17BQwen403B1447.2189618~5.7Aug 2, 2026leaderboard35o3OpenAI1447.0532181—Aug 2, 2026leaderboard36DeepSeek V4 ProDeepSeek1.6T1445.32071661~1.4Aug 2, 2026leaderboard37GLM 5Z.AI754B1442.3273751~3Aug 2, 2026leaderboard38MiMo-V2.5Xiaomi311B1441.48876435~7.4Aug 2, 2026leaderboard39Gemini 2.5 ProGoogle1440.8777288—Aug 2, 2026leaderboard40Kimi K2.5Moonshot AI1.1T1440.6502720~2.2Aug 2, 2026leaderboard41GPT-5.4 MiniOpenAI1439.7514via High640—Aug 2, 2026leaderboard42MiniMax M3MiniMax427B1439.51441499~5.3Aug 2, 2026leaderboard43Gemini 3.1 Flash Lite PreviewGoogle1439.00791919—Aug 2, 2026leaderboard44Qwen3 MaxQwen1438.3285369—Aug 2, 2026leaderboard45Claude Opus 4.1Anthropic1433.463719—Aug 2, 2026leaderboard46Mistral Medium 3.5Mistral AI128B1430.637191~18Aug 2, 2026leaderboard47DeepSeek V3.2DeepSeek685B1428.48294596~3.3Aug 2, 2026leaderboard48Qwen3.5-27BQwen27.8B1428.2094916~81Aug 2, 2026leaderboard49GLM 4.7Z.AI358B1428.1583816~6.3Aug 2, 2026leaderboard50Hy3 previewTencent299B1428.04586800~7.6Aug 2, 2026leaderboard51Claude Sonnet 4.5Anthropic1427.775895—Aug 2, 2026leaderboard52DeepSeek V4 FlashDeepSeek291B1426.60568087~7.8Aug 2, 2026leaderboard53GPT-5.3 ChatOpenAI1425.996102—Aug 2, 2026leaderboard54GPT-5.4 NanoOpenAI1423.7457via High2278—Aug 2, 2026leaderboard55GPT-5.1OpenAI1423.2971285—Aug 2, 2026leaderboard56MiniMax M2.7MiniMax229B1423.07851482~9.9Aug 2, 2026leaderboard57Qwen3.5-122B-A10BQwen125B1422.9039684~18Aug 2, 2026leaderboard58Gemini 3.5 Flash LiteGoogle1420.3961136—Aug 2, 2026leaderboard59GLM 4.6Z.AI357B1420.0742811~6.3Aug 2, 2026leaderboard60Grok 4.3xAI1419.167568—Aug 2, 2026leaderboard61Qwen3 235B A22B Instruct 2507Qwen235B1418.34132579~9.6Aug 2, 2026leaderboard62DeepSeek V3.2 ExpDeepSeek685B1417.81863458~3.3Aug 2, 2026leaderboard63Qwen3 Next 80B A3B InstructQwen81.3B1417.53961817~28Aug 2, 2026leaderboard64o4 MiniOpenAI1415.565322—Aug 2, 2026leaderboard65GLM 4.5Z.AI358B1412.9736642~6.3Aug 2, 2026leaderboard66R1 0528DeepSeek685B1411.7154657~3.3Aug 2, 2026leaderboard67Qwen3 VL 235B A22B InstructQwen236B1410.42251603~9.5Aug 2, 2026leaderboard68o1OpenAI1408.894423—Aug 2, 2026leaderboard69Step 3.5 FlashStepFun199B1407.30694691~11Aug 2, 2026leaderboard70Gemini 2.5 FlashGoogle1405.51221124—Aug 2, 2026leaderboard71o3 Mini HighOpenAI1405.3327via High319—Aug 2, 2026leaderboard72GPT-5 MiniOpenAI1405.2541via High1405—Aug 2, 2026leaderboard73Qwen3 VL 235B A22B ThinkingQwen236B1404.1609via Thinking540~9.4Aug 2, 2026leaderboard74Qwen3.5-FlashQwen1403.82075399—Aug 2, 2026leaderboard75Claude Opus 4Anthropic1403.527919—Aug 2, 2026leaderboard76Mistral Large 3 2512Mistral AI1402.2942935—Aug 2, 2026leaderboard77Qwen3.5-35B-A3BQwen36B1400.28411400~62Aug 2, 2026leaderboard78Claude Haiku 4.5Anthropic1398.8869280—Aug 2, 2026leaderboard79Qwen3 32BQwen32.8B1398.7214995~68Aug 2, 2026leaderboard80Qwen3 235B A22B Thinking 2507Qwen235B1397.2464935~9.4Aug 2, 2026leaderboard81MiniMax M2.5MiniMax229B1396.26741551~9.7Aug 2, 2026leaderboard82DeepSeek V3.1 TerminusDeepSeek685B1396.11191470~3.2Aug 2, 2026leaderboard83GLM 4.5 AirZ.AI111B1389.58541635~20Aug 2, 2026leaderboard84Qwen3 Next 80B A3B ThinkingQwen81.3B1389.3319via Thinking1781~27Aug 2, 2026leaderboard85Trinity Large ThinkingArcee AI399B1384.1305via Thinking1730~5.5Aug 2, 2026leaderboard86o3 MiniOpenAI1381.5397314—Aug 2, 2026leaderboard87gpt-oss-120bOpenAI120B1381.30068125~18Aug 2, 2026leaderboard88Qwen3 30B A3B Instruct 2507Qwen30.5B1380.19457149~72Aug 2, 2026leaderboard89Qwen3 Coder 480B A35BQwen480B1376.44531376~4.5Aug 2, 2026leaderboard90Nemotron 3 SuperNVIDIA124B1375.5983439~18Aug 2, 2026leaderboard91GPT-4.1OpenAI1373.3339172—Aug 2, 2026leaderboard92MiniMax M1MiniMax1371.3322623—Aug 2, 2026leaderboard93DeepSeek V3 0324DeepSeek685B1369.58931522~3.2Aug 2, 2026leaderboard94GLM 4.7 FlashZ.AI31.2B1365.0223341372Aug 2, 2026leaderboard95GLM 4.5VZ.AI108B1360.6764756~20Aug 2, 2026leaderboard96MiniMax M2MiniMax229B1354.75771328~9.4Aug 2, 2026leaderboard97GPT-4.1 MiniOpenAI1353.6827846—Aug 2, 2026leaderboard98Ling-2.6-flashinclusionAI107B1352.72745091~20Aug 2, 2026leaderboard99Nemotron 3 Nano 30B A3BNVIDIA31.6B1351.76946759~68Aug 2, 2026leaderboard100Claude Sonnet 5Anthropic1351.1821135—Aug 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.