Leaderboards

Benchmarks

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

Reasoning280Coding266Math174Agentic110Chat515Speech654
Arena Creative Writing143Arena Instruction Following143Arena Text (overall)143IFEval16LiveBench Instruction Following35LiveBench Language35

IFEval

% instruction-following accuracyhigher is better
Provenanceindependent leaderboard16 modelsscores Aug 3, 2026
1Llama 3.3 70B InstructMeta70.6B90281~2Aug 3, 2026leaderboard2Llama 3.1 70B InstructMeta70.6B86.7217~1.9Aug 3, 2026leaderboard3Qwen2.5 72B InstructQwen72.7B84.8212~1.8Aug 3, 2026leaderboard4Mistral LargeMistral AI8414Aug 3, 2026leaderboard5Gemma 2 27BGoogle27.2B79.8123~4.7Aug 3, 2026leaderboard6Command R7B (12-2024)Cohere77.1514Aug 3, 2026leaderboard7Hermes 3 70B InstructNous Research70.6B76.6109~1.7Aug 3, 2026leaderboard8Qwen2.5 7B InstructQwen7.6B72.81040~15Aug 3, 2026leaderboard9Qwen2.5 Coder 32B InstructQwen32.8B72.773~3.5Aug 3, 2026leaderboard10Llama 3.2 3B InstructMeta3.2B721440~36Aug 3, 2026leaderboard11Mixtral 8x22B InstructMistral AI141B71.812~0.81Aug 3, 2026leaderboard12Mistral NemoMistral AI12.2B63.82127~8.3Aug 3, 2026leaderboard13Llama 3.2 1B InstructMeta1.2B54.82740~72Aug 3, 2026leaderboard14Llama 3.1 8B InstructMeta8B49.41235~9.8Aug 3, 2026leaderboard15R1 Distill Llama 70BDeepSeek70.6B43.454~0.98Aug 3, 2026leaderboard16Phi 4Microsoft14.7B5.9421Aug 3, 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.