LLM Stats · GitHub archive
Indexed results from LLM Stats · GitHub archive. Each ranking retains its own metric and evaluation conditions.
MMLU · Raw
llmStatsArchive · LLM Stats · GitHub archive · MMLU · Raw · score · Higher is better
MBPP · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · MBPP · Normalized (reported) · score · Higher is better
GPQA · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · GPQA · Normalized (reported) · score · Higher is better
MATH-500 · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · MATH-500 · Normalized (reported) · score · Higher is better
- DeepSeek: R1deepseek0.959
- DeepSeek: DeepSeek-R1-Distill-Llama-70Bdeepseek0.945
- deepseek-ai/DeepSeek-R1-Distill-Qwen-32Bdeepseek0.943
- DeepSeek: DeepSeek V3 0324deepseek0.94
- DeepSeek: DeepSeek-R1-Distill-Qwen-14Bdeepseek0.939
- DeepSeek: DeepSeek-R1-Distill-Qwen-7Bdeepseek0.928
- DeepSeek: DeepSeek-R1-Distill-Llama-8Bdeepseek0.891
- deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5Bdeepseek-ai0.839
ComplexFuncBench · Raw
llmStatsArchive · LLM Stats · GitHub archive · ComplexFuncBench · Raw · score · Higher is better
AIME 2025 · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · AIME 2025 · Normalized (reported) · score · Higher is better
HellaSwag · Raw
llmStatsArchive · LLM Stats · GitHub archive · HellaSwag · Raw · score · Higher is better
- Qwen/Qwen2-72Bqwen0.876
GSM8k · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · GSM8k · Normalized (reported) · score · Higher is better
Android Control Low_EM · Raw
llmStatsArchive · LLM Stats · GitHub archive · Android Control Low_EM · Raw · score · Higher is better
MMMU · Raw
llmStatsArchive · LLM Stats · GitHub archive · MMMU · Raw · score · Higher is better
AutoLogi · Raw
llmStatsArchive · LLM Stats · GitHub archive · AutoLogi · Raw · score · Higher is better
AITZ_EM · Normalized (reported)
llmStatsArchive · LLM Stats · GitHub archive · AITZ_EM · Normalized (reported) · score · Higher is better
Models
LoadingBenchmarks
Explore every benchmark
LoadingThese charts contain model-linked results. The charts above also include unlinked source profiles. Compare scores within each evaluation. Model quality, usage, transparency and system performance measure different things.