LLMap
Leaderboards

Benchmarks

Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.

Reasoning280Coding266Math174Agentic110Chat515Speech654
Arena Agent (IPS)36LiveBench Agentic Coding35SWE-bench Verified39

Arena Agent (IPS)

Inverse-propensity score from live agent sessions — neither an Elo nor a percentagehigher is better
Provenanceindependent leaderboard36 modelsscores Jul 28, 2026
methodology ↗source ↗
1Claude Fable 5Anthropic0.1258via High0—Jul 28, 2026leaderboard2Claude Opus 5Anthropic0.1188via Max0—Jul 28, 2026leaderboard3GPT-5.6 SolOpenAI0.1002via xHigh0—Jul 28, 2026leaderboard4Kimi K3Moonshot AI2.8T0.0991via Max0~0.000057Jul 28, 2026leaderboard5Claude Sonnet 5Anthropic0.0857via High0—Jul 28, 2026leaderboard6Claude Opus 4.7Anthropic0.07370—Jul 28, 2026leaderboard7GLM 5.2Z.AI753B0.0708via Max0~0.00015Jul 28, 2026leaderboard8Claude Opus 4.6Anthropic0.06470—Jul 28, 2026leaderboard9Grok 4.5xAI0.05540—Jul 28, 2026leaderboard10GPT-5.5OpenAI0.05530—Jul 28, 2026leaderboard11GPT-5.4OpenAI0.0511via High0—Jul 28, 2026leaderboard12GPT-5.6 TerraOpenAI0.0353via xHigh0—Jul 28, 2026leaderboard13GPT-5.6 LunaOpenAI0.0347via xHigh0—Jul 28, 2026leaderboard14Claude Opus 4.8Anthropic0.03340—Jul 28, 2026leaderboard15Claude Sonnet 4.6Anthropic0.03180—Jul 28, 2026leaderboard16Muse Spark 1.1Meta0.00750—Jul 28, 2026leaderboard17GLM 5.1Z.AI754B0.00440~0.0000093Jul 28, 2026leaderboard18Kimi K2.7 CodeMoonshot AI1.1T0.00440~0.0000066Jul 28, 2026leaderboard19Gemini 3.1 Pro Preview Custom ToolsGoogle-0.0052-0—Jul 28, 2026leaderboard20Qwen3.7 MaxQwen-0.0053-0—Jul 28, 2026leaderboard21DeepSeek V4 ProDeepSeek1.6T-0.0078-0~-0.0000077Jul 28, 2026leaderboard22Gemini 3.5 FlashGoogle-0.0103via High-0—Jul 28, 2026leaderboard23Hy3Tencent299B-0.0103-0~-0.000055Jul 28, 2026leaderboard24Kimi K2.6Moonshot AI1.1T-0.0109-0~-0.000016Jul 28, 2026leaderboard25Gemini 3.6 FlashGoogle-0.0159-0—Jul 28, 2026leaderboard26Qwen3.7 PlusQwen-0.0165-0—Jul 28, 2026leaderboard27MiMo-V2.5-ProXiaomi1T-0.0252-0~-0.000039Jul 28, 2026leaderboard28MiniMax M3MiniMax427B-0.0255-0~-0.000095Jul 28, 2026leaderboard29DeepSeek V4 FlashDeepSeek291B-0.0296-0~-0.00016Jul 28, 2026leaderboard30InklingThinking Machines952B-0.0571-0~-0.000095Jul 28, 2026leaderboard31Grok Build 0.1xAI-0.0799-0—Jul 28, 2026leaderboard32Gemini 3.5 Flash LiteGoogle-0.102-0—Jul 28, 2026leaderboard33MiniMax M2.7MiniMax229B-0.1159-0~-0.0008Jul 28, 2026leaderboard34Nemotron 3 UltraNVIDIA561B-0.1314-0~-0.00037Jul 28, 2026leaderboard35Grok 4.3xAI-0.1445-0—Jul 28, 2026leaderboard36Gemma 4 31BGoogle32.7B-0.161-0~-0.0078Jul 28, 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.