Leaderboards
Benchmarks
Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.
Arena Code (WebDev)
Arena Score from human votes on web-app building tasks — not a percentagehigher is betterProvenanceindependent leaderboard74 modelsscores Aug 2, 2026
1Claude Opus 5Anthropic1704.6725via Max68—Aug 2, 2026leaderboard2Kimi K3Moonshot AI2.8T1675.5597via Max120~0.96Aug 2, 2026leaderboard3Claude Fable 5Anthropic1630.023633—Aug 2, 2026leaderboard4GPT-5.6 SolOpenAI1620.2674via codex-harness108—Aug 2, 2026leaderboard5GLM 5.2Z.AI753B1586.2924via Max1058~3.3Aug 2, 2026leaderboard6DeepSeek V4 FlashDeepSeek291B1576.5385via High8937~8.6Aug 2, 2026leaderboard7Claude Opus 4.7Anthropic1560.631362—Aug 2, 2026leaderboard8Grok 4.5xAI1549.2377258—Aug 2, 2026leaderboard9Claude Sonnet 5Anthropic1541.8892via High154—Aug 2, 2026leaderboard10Claude Opus 4.8Anthropic1538.83862—Aug 2, 2026leaderboard11Claude Opus 4.6Anthropic1537.92662—Aug 2, 2026leaderboard12Muse Spark 1.1Meta1535.9479361—Aug 2, 2026leaderboard13Gemini 3.6 FlashGoogle1532.5354409—Aug 2, 2026leaderboard14GPT-5.6 LunaOpenAI1522.941via codex-harness1154—Aug 2, 2026leaderboard15Claude Sonnet 4.6Anthropic1522.6366102—Aug 2, 2026leaderboard16GPT-5.6 TerraOpenAI1521.6062via codex-harness507—Aug 2, 2026leaderboard17Qwen3.7 MaxQwen1516.9691405—Aug 2, 2026leaderboard18Hy3Tencent299B1516.52542872~8Aug 2, 2026leaderboard19GLM 5.1Z.AI754B1516.1626507~3.2Aug 2, 2026leaderboard20Kimi K2.6Moonshot AI1.1T1509.5097609~2.3Aug 2, 2026leaderboard21GPT-5.5OpenAI1508.9407via codex-harness151—Aug 2, 2026leaderboard22MiniMax M3MiniMax427B1490.74431553~5.5Aug 2, 2026leaderboard23Gemini 3.5 FlashGoogle1486.184via Medium330—Aug 2, 2026leaderboard24Qwen3.6 Max PreviewQwen1478.1786240—Aug 2, 2026leaderboard25MiMo-V2.5-ProXiaomi1T1474.01782118~2.3Aug 2, 2026leaderboard26Kimi K2.7 CodeMoonshot AI1.1T1472.5712421~2.2Aug 2, 2026leaderboard27Claude Opus 4.5Anthropic1466.990359—Aug 2, 2026leaderboard28Qwen3.6 PlusQwen1457.8583748—Aug 2, 2026leaderboard29Gemini 3.5 Flash LiteGoogle1451.87451161—Aug 2, 2026leaderboard30Gemini 3.1 Pro Preview Custom ToolsGoogle1446.8335241—Aug 2, 2026leaderboard31DeepSeek V4 ProDeepSeek1.6T1446.00121662~1.4Aug 2, 2026leaderboard32Kimi K2 ThinkingMoonshot AI1.1T1435.7698via Thinking574~2.2Aug 2, 2026leaderboard33MiMo-V2.5Xiaomi311B1435.49666408~7.3Aug 2, 2026leaderboard34GLM 5Z.AI754B1434.742747~3Aug 2, 2026leaderboard35GLM 4.7Z.AI358B1433.4636819~6.3Aug 2, 2026leaderboard36GPT-5.2 ChatOpenAI1419.1946203—Aug 2, 2026leaderboard37InklingThinking Machines952B1411.4606349~2.4Aug 2, 2026leaderboard38GPT-5.3-CodexOpenAI1408.8955via codex-harness101—Aug 2, 2026leaderboard39Kimi K2.5Moonshot AI1.1T1404.7669702~2.1Aug 2, 2026leaderboard40Qwen3.5 397B A17BQwen403B1399.5863598~5.5Aug 2, 2026leaderboard41GLM 5V TurboZ.AI1399.349350—Aug 2, 2026leaderboard42GPT-5.4 MiniOpenAI1398.6596via High622—Aug 2, 2026leaderboard43MiniMax M2.7MiniMax229B1398.49241457~9.7Aug 2, 2026leaderboard44GPT-5.4OpenAI1389.2851185—Aug 2, 2026leaderboard45Claude Opus 4.1Anthropic1388.740919—Aug 2, 2026leaderboard46MiniMax M2.5MiniMax229B1386.26611540~9.6Aug 2, 2026leaderboard47Claude Sonnet 4.5Anthropic1385.125392—Aug 2, 2026leaderboard48Gemma 4 26B A4BGoogle26.5B1366.58884555~82Aug 2, 2026leaderboard49Gemma 4 31BGoogle32.7B1363.39224010~66Aug 2, 2026leaderboard50Qwen3.5-122B-A10BQwen125B1359.5039654~17Aug 2, 2026leaderboard51Qwen3.5-27BQwen27.8B1357.616870~77Aug 2, 2026leaderboard52Grok 4.3xAI1356.9268543—Aug 2, 2026leaderboard53Hy3 previewTencent299B1356.77126461~7.2Aug 2, 2026leaderboard54Laguna M.1poolside226B1348.81443372~9.5Aug 2, 2026leaderboard55GPT-5.1OpenAI1341.2741268—Aug 2, 2026leaderboard56GLM 4.6Z.AI357B1339.0301765~6Aug 2, 2026leaderboard57GPT-5.2-CodexOpenAI1338.403896—Aug 2, 2026leaderboard58GPT-5.1-CodexOpenAI1336.0713134—Aug 2, 2026leaderboard59Claude Haiku 4.5Anthropic1325.0375265—Aug 2, 2026leaderboard60DeepSeek V3.2DeepSeek685B1323.3844258~3.1Aug 2, 2026leaderboard61MiniMax M2MiniMax229B1296.6921271~9Aug 2, 2026leaderboard62Qwen3 Coder 480B A35BQwen480B1272.21571272~4.2Aug 2, 2026leaderboard63DeepSeek V3.2 ExpDeepSeek685B1271.77893102~2.9Aug 2, 2026leaderboard64Mistral Medium 3.5Mistral AI128B1267.0675169~16Aug 2, 2026leaderboard65Gemini 3.1 Flash Lite PreviewGoogle1256.22581675—Aug 2, 2026leaderboard66Qwen3.5-35B-A3BQwen36B1250.46041250~55Aug 2, 2026leaderboard67GPT-5.1-Codex-MiniOpenAI1244.2131622—Aug 2, 2026leaderboard68Trinity Large ThinkingArcee AI399B1238.9321via Thinking1549~4.9Aug 2, 2026leaderboard69Qwen3.5-FlashQwen1237.95844761—Aug 2, 2026leaderboard70Mistral Large 3 2512Mistral AI1230.0154820—Aug 2, 2026leaderboard71Gemini 2.5 ProGoogle1224.2315245—Aug 2, 2026leaderboard72Granite 4.1 8BIBM8.8B1193.975611940~220Aug 2, 2026leaderboard73Devstral 2 2512Mistral AI125B1193.8917597~15Aug 2, 2026leaderboard74Mercury 2Inception1166.34341555—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.