Scores updated · 20 tests both models report · How we compare
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results20 tests
- C-EvalDeepSeek-V4-Pro-Base by 20.693.172.5+20.6
- CMMLUDeepSeek-V4-Pro-Base by 17.190.873.7+17.1
- MATH (EM)DeepSeek-V4-Pro-Base by 15.564.549+15.5
- MGSM (EM)DeepSeek-V4-Pro-Base by 14.584.469.9+14.5
- CMath (EM)DeepSeek-V4-Pro-Base by 13.690.977.3+13.6
- HumanEvalLlama 3.1 Instruct 405B by 12.276.889+12.2
- MATHLlama 3.1 Instruct 405B by 9.364.573.8+9.3
- GSM8K (EM)DeepSeek-V4-Pro-Base by 9.192.683.5+9.1
- BBHDeepSeek-V4-Pro-Base by 4.687.582.9+4.6
- BBH (EM)DeepSeek-V4-Pro-Base by 4.687.582.9+4.6
- MMLU-ReduxDeepSeek-V4-Pro-Base by 4.690.886.2+4.6
- GSM8KLlama 3.1 Instruct 405B by 4.292.696.8+4.2
- DROPDeepSeek-V4-Pro-Base by 3.988.784.8+3.9
- WinoGrandeLlama 3.1 Instruct 405B by 3.781.585.2+3.7
- TriviaQA (EM)DeepSeek-V4-Pro-Base by 2.985.682.7+2.9
- DROP (F1)DeepSeek-V4-Pro-Base by 2.788.786+2.7
- CLUEWSCDeepSeek-V4-Pro-Base by 2.285.283+2.2
- MMLUDeepSeek-V4-Pro-Base by 1.590.188.6+1.5
- HellaSwagLlama 3.1 Instruct 405B by 1.28889.2+1.2
- MMLU-Protie73.573.3tie
Questions people ask
How do you compare the two?
We use the 20 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.