DeepSeek-R1-Zero is the stronger all-rounder.
Scores updated · 15 tests both models report · How we compare
Where each one wins
Tests won in each of the two 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 and reasoning 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 DeepSeek-R1-Zero pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+21.8points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+18.7points ahead
Where Llama 3.1 Instruct 405B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 15 tests both models report. The winning score is in its model's colour; marks a score checked independently.
Other results13 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.
- CodeforcesDeepSeek-R1-Zero by 1419 rating points25.31444+1419 rating
- AIME 2024DeepSeek-R1-Zero by 54.623.377.9+54.6
- IFEvalLlama 3.1 Instruct 405B by 4288.646.6+42
- LiveCodeBenchDeepSeek-R1-Zero by 22.327.750+22.3
- MATH-500 (EM)DeepSeek-R1-Zero by 22.173.895.9+22.1
- Chinese SimpleQA (C-SimpleQA)DeepSeek-R1-Zero by 1650.466.4+16
- SimpleQADeepSeek-R1-Zero by 13.217.130.3+13.2
- FRAMES (Acc.)DeepSeek-R1-Zero by 12.37082.3+12.3
- CLUEWSCDeepSeek-R1-Zero by 10.18393.1+10.1
- MMLU-ProLlama 3.1 Instruct 405B by 4.473.368.9+4.4
- MMLU-Reduxtie86.285.6tie
- DROP (3-shot F1)tie88.789.1tie
- MMLUtie88.688.8tie
Questions people ask
Which is better, Llama 3.1 Instruct 405B or DeepSeek-R1-Zero?
DeepSeek-R1-Zero wins all two areas where both have results: coding and reasoning. Llama 3.1 Instruct 405B wins none.
Which is better for coding?
DeepSeek-R1-Zero. It wins the one coding test both models report.
How do you compare the two?
We use the 15 benchmark tests both models have published scores on. The verdict counts the 2 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 13 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.