The two are evenly matched.
Scores updated · 22 tests both models report · How we compare
Where each one wins
Tests won in each of the one area where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
Reasoning rests on a single test.
Every test, side by side
All 22 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results21 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- MultiPL-E HumanEvalLlama 3.1 Instruct 405B by 29.345.975.2+29.3
- Chinese SimpleQA (C-SimpleQA)DeepSeek-V3.1-Base by 20.570.950.4+20.5
- C-EvalDeepSeek-V3.1-Base by 17.59072.5+17.5
- CRUXEval-O (output prediction)DeepSeek-V3.1-Base by 16.576.459.9+16.5
- HumanEvalLlama 3.1 Instruct 405B by 16.572.589+16.5
- CMMLUDeepSeek-V3.1-Base by 15.188.873.7+15.1
- MMLU-ProLlama 3.1 Instruct 405B by 14.558.873.3+14.5
- MultiPL-E MBPPLlama 3.1 Instruct 405B by 13.252.565.7+13.2
- MATHLlama 3.1 Instruct 405B by 11.262.673.8+11.2
- SimpleQADeepSeek-V3.1-Base by 9.226.317.1+9.2
- MBPPDeepSeek-V3.1-Base by 6.274.668.4+6.2
- GSM8KLlama 3.1 Instruct 405B by 5.491.496.8+5.4
- BBHDeepSeek-V3.1-Base by 5.388.282.9+5.3
- MMLU-ReduxDeepSeek-V3.1-Base by 3.89086.2+3.8
- CRUXEval-I (input prediction)DeepSeek-V3.1-Base by 3.662.158.5+3.6
- DROPDeepSeek-V3.1-Base by 1.586.384.8+1.5
- ARC-ChallengeLlama 3.1 Instruct 405B by 1.395.696.9+1.3
- MMLULlama 3.1 Instruct 405B by 1.287.488.6+1.2
- MBPP+ (EvalPlus-augmented)tie72.273tie
- WinoGrandetie85.985.2tie
- HellaSwagtie89.289.2tie
Questions people ask
Which is better, DeepSeek-V3.1-Base or Llama 3.1 Instruct 405B?
Neither wins more tests than the other in any of the one area where both have results.
How do you compare the two?
We use the 22 benchmark tests both models have published scores on. The verdict counts the 1 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 21 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.