Scores updated · 23 tests both models report · How we compare
Every test, side by side
All 23 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results23 tests
- FACTS Parametric (EM)DeepSeek-V4-Pro-Base by 35.527.162.6+35.5
- Simple-QA verified (EM)DeepSeek-V4-Pro-Base by 26.928.355.2+26.9
- HumanEvalDeepSeek-V4-Pro-Base by 1462.876.8+14
- MultiLoKo (EM)DeepSeek-V4-Pro-Base by 12.438.751.1+12.4
- LongBench-V2 (EM)DeepSeek-V4-Pro-Base by 11.340.251.5+11.3
- SuperGPQA (EM)DeepSeek-V4-Pro-Base by 8.94553.9+8.9
- MMLU-ProDeepSeek-V4-Pro-Base by 865.573.5+8
- BigCodeBenchDeepSeek-V3.2-Base by 4.763.959.2+4.7
- AGIEval (EM)DeepSeek-V4-Pro-Base by 4.380.184.4+4.3
- MATH (EM)DeepSeek-V4-Pro-Base by 460.564.5+4
- MMLU-ReduxDeepSeek-V4-Pro-Base by 3.387.590.8+3.3
- MGSM (EM)DeepSeek-V4-Pro-Base by 3.181.384.4+3.1
- WinoGrande (EM)DeepSeek-V4-Pro-Base by 2.678.981.5+2.6
- MMMLU (EM)DeepSeek-V4-Pro-Base by 2.487.990.3+2.4
- MMLUDeepSeek-V4-Pro-Base by 2.387.890.1+2.3
- TriviaQA (EM)DeepSeek-V4-Pro-Base by 2.383.385.6+2.3
- CMMLU (EM)DeepSeek-V4-Pro-Base by 1.988.990.8+1.9
- CLUEWSCDeepSeek-V4-Pro-Base by 1.783.585.2+1.7
- CMath (EM)DeepSeek-V3.2-Base by 1.792.690.9+1.7
- HellaSwag (EM)DeepSeek-V4-Pro-Base by 1.686.488+1.6
- GSM8K (EM)DeepSeek-V4-Pro-Base by 1.591.192.6+1.5
- DROP (F1)tie88.288.7tie
- BBH (EM)tie87.687.5tie
Questions people ask
How do you compare the two?
We use the 23 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.