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
GPQA Diamond
% correct, graduate-level science questionshigher is betterProvenanceindependent leaderboard16 modelsscores Aug 3, 2026
1Mistral LargeMistral AI24.94—Aug 3, 2026leaderboard2Phi 4Microsoft14.7B20.81492Aug 3, 2026leaderboard3Gemma 2 27BGoogle27.2B16.726~0.97Aug 3, 2026leaderboard4Mixtral 8x22B InstructMistral AI141B16.43~0.18Aug 3, 2026leaderboard5Hermes 3 70B InstructNous Research70.6B14.921~0.33Aug 3, 2026leaderboard6Qwen2.5 72B InstructQwen72.7B14.536~0.32Aug 3, 2026leaderboard7Llama 3.1 70B InstructMeta70.6B14.235~0.32Aug 3, 2026leaderboard8Qwen2.5 Coder 32B InstructQwen32.8B13.213~0.64Aug 3, 2026leaderboard9Llama 3.3 70B InstructMeta70.6B10.533~0.24Aug 3, 2026leaderboard10Command R7B (12-2024)Cohere7.852—Aug 3, 2026leaderboard11Qwen2.5 7B InstructQwen7.6B6.593~1.4Aug 3, 2026leaderboard12Mistral NemoMistral AI12.2B5.4180~0.7Aug 3, 2026leaderboard13Llama 3.2 3B InstructMeta3.2B2.856~1.4Aug 3, 2026leaderboard14Llama 3.1 8B InstructMeta8B2.563~0.5Aug 3, 2026leaderboard15R1 Distill Llama 70BDeepSeek70.6B23~0.045Aug 3, 2026leaderboard16Llama 3.2 1B InstructMeta1.2B1.680~2.1Aug 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.