Llama 3.1 Instruct 405B is the stronger all-rounder.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the three areas 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.
Coding, reasoning and long documents rest on a single test each.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Llama 3.1 Instruct 405B pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+9.2points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+7.7points ahead
Where DeepSeek-V2.5 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingLlama 3.1 Instruct 405B
- SWE-bench VerifiedLlama 3.1 Instruct 405B by 7.716.824.5+7.7
ReasoningLlama 3.1 Instruct 405B
- GPQA DiamondLlama 3.1 Instruct 405B by 9.242.351.5+9.2
Other results25 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.
- AlignBenchDeepSeek-V2.5 by 74.480.46+74.4
- LiveCodeBenchDeepSeek-V2.5 by 14.141.827.7+14.1
- AlpacaEval-2.0DeepSeek-V2.5 by 11.250.539.3+11.2
- CodeforcesDeepSeek-V2.5 by 10.335.625.3+10.3
- MMLULlama 3.1 Instruct 405B by 8.280.488.6+8.2
- IFEvalLlama 3.1 Instruct 405B by 880.688.6+8
- Aider-Edit (Acc.)DeepSeek-V2.5 by 7.771.663.9+7.7
- CLUEWSCDeepSeek-V2.5 by 7.490.483+7.4
- MMLU-ProLlama 3.1 Instruct 405B by 7.166.273.3+7.1
- Arena HardDeepSeek-V2.5 by 6.976.269.3+6.9
- SimpleQALlama 3.1 Instruct 405B by 6.910.217.1+6.9
- AIME 2024Llama 3.1 Instruct 405B by 6.616.723.3+6.6
- MMLU-ReduxLlama 3.1 Instruct 405B by 5.980.386.2+5.9
- FRAMES (Acc.)Llama 3.1 Instruct 405B by 4.665.470+4.6
- Chinese SimpleQA (C-SimpleQA)DeepSeek-V2.5 by 3.754.150.4+3.7
- HumanEval-Mul (Pass@1)Llama 3.1 Instruct 405B by 3.473.877.2+3.4
- GSM8KLlama 3.1 Instruct 405B by 1.795.196.8+1.7
- BBHDeepSeek-V2.5 by 1.484.382.9+1.4
- HellaSwagDeepSeek-V2.5 by 1.190.389.2+1.1
- DROP (3-shot F1)tie87.888.7tie
- MATHtie74.773.8tie
- MATH-500 (EM)tie74.773.8tie
- AA Intelligencetie6.67.3tie
- MT-Benchtie98.5tie
- HumanEvaltie8989tie
Questions people ask
Which is better, DeepSeek-V2.5 or Llama 3.1 Instruct 405B?
Llama 3.1 Instruct 405B wins two of the three areas where both have results: coding and reasoning. DeepSeek-V2.5 wins none. They are level on long documents.
Which is better for coding?
Llama 3.1 Instruct 405B. It wins the one coding test both models report.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 3 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 25 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.