NousResearch/

eval-Cogito-v2-preview-405B-reasoning

	cogito-405b-thinking Evaluation Results




	
	Summary




	

Benchmark Score Metric Samples Overlong rate

aime24 0.408 math_pass@1:64_samples 64 20.3%

aime25 0.327 math_pass@1:64_samples 64 15.5%

arenahard 0.910 eval/overall_winrate 500 0.0%

bbh_generative 0.893 extractive_match 1 1.3%

creative-writing-v3 0.674 creative_writing_score 96 0.0%

drop_generative_nous 0.871 drop_acc 1 0.3%

eqbench3 0.672 eqbench_score 135 0.0%

gpqa_diamond 0.682… See the full description on the dataset page: https://huggingface.co/datasets/NousResearch/eval-Cogito-v2-preview-405B-reasoning.