Scores updated · 16 tests both models report · How we compare
Every test, side by side
All 16 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results16 tests
- Chinese SimpleQA (C-SimpleQA)DeepSeek-V3-Base by 18.368.750.4+18.3
- C-EvalDeepSeek-V3-Base by 17.990.372.5+17.9
- MATHLlama 3.1 Instruct 405B by 14.459.473.8+14.4
- CMMLUDeepSeek-V3-Base by 14.287.973.7+14.2
- CRUXEval-O (output prediction)DeepSeek-V3-Base by 11.17159.9+11.1
- CRUXEval-I (input prediction)DeepSeek-V3-Base by 9.367.858.5+9.3
- MMLU-ProLlama 3.1 Instruct 405B by 9.36473.3+9.3
- SimpleQADeepSeek-V3-Base by 9.126.117.1+9.1
- GSM8KLlama 3.1 Instruct 405B by 8.688.296.8+8.6
- MMMLUDeepSeek-V3-Base by 5.779.573.8+5.7
- MBPP+ (EvalPlus-augmented)DeepSeek-V3-Base by 2.575.573+2.5
- ARC-ChallengeLlama 3.1 Instruct 405B by 2.294.796.9+2.2
- DROPDeepSeek-V3-Base by 1.786.584.8+1.7
- PIQALlama 3.1 Instruct 405B by 1.784.285.9+1.7
- MMLUtie87.788.6tie
- MMLU-Reduxtie86.886.2tie
Questions people ask
How do you compare the two?
We use the 16 benchmark tests both models have published scores on. None of them is in the eight capability areas we count, so this page lists them without a verdict. Each score is the one shown on the model's own page.