DeepSeek-R1-Distill-Llama-70B vs DeepSeek-R1-Distill-Qwen-32B
Wins 2 of 4 areas
Long documents · Following instructions
Wins 2 of 4 areas
Coding · Reasoning
The two are evenly matched.DeepSeek-R1-Distill-Llama-70B is better at long documents and following instructions; DeepSeek-R1-Distill-Qwen-32B at reasoning and coding.
Scores updated · 12 tests both models report · How we compare
Where each one wins
Tests won in each of the four 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.
- Long documentsFinding answers in very long texts10DeepSeek-R1-Distill-Llama-70B1 of 1 test
- Following instructionsDoing exactly what it is asked10DeepSeek-R1-Distill-Llama-70B1 of 1 test
- ReasoningHard problems that need careful thinking01DeepSeek-R1-Distill-Qwen-32B1 of 2 tests · 1 tie
- CodingWriting and fixing software01DeepSeek-R1-Distill-Qwen-32B1 of 1 test
Coding, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
DeepSeek-R1-Distill-Qwen-32B costs 68% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where DeepSeek-R1-Distill-Llama-70B pulls ahead
Where DeepSeek-R1-Distill-Qwen-32B pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+21.3points ahead
- Code for real scientific research problemsSciCode+6.3points ahead
Every test, side by side
All 12 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-R1-Distill-Qwen-32B
- SciCodeDeepSeek-R1-Distill-Qwen-32B by 6.331.337.6+6.3
ReasoningDeepSeek-R1-Distill-Qwen-32B
- GPQA DiamondDeepSeek-R1-Distill-Qwen-32B by 21.340.261.5+21.3
- Humanity's Last Examtie5.14.6tie
Long documentsDeepSeek-R1-Distill-Llama-70B
- AA-LCRDeepSeek-R1-Distill-Llama-70B by 1.3108.7+1.3
Following instructionsDeepSeek-R1-Distill-Llama-70B
- IFBenchDeepSeek-R1-Distill-Llama-70B by 4.727.622.9+4.7
Other results7 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.
- Codeforces (Rating)DeepSeek-R1-Distill-Qwen-32B by 58 rating points16331691+58 rating
- AIME 2025DeepSeek-R1-Distill-Qwen-32B by 55560+5
- AIME 2024DeepSeek-R1-Distill-Qwen-32B by 2.67072.6+2.6
- AA Intelligencetie7.98.4tie
- LiveCodeBenchtie57.557.2tie
- MATH-500 (EM)tie94.594.3tie
- HMMT Feb. 2025tie33.333.3tie
Questions people ask
Which is better, DeepSeek-R1-Distill-Llama-70B or DeepSeek-R1-Distill-Qwen-32B?
DeepSeek-R1-Distill-Llama-70B and DeepSeek-R1-Distill-Qwen-32B each win two of the four areas where both have results. DeepSeek-R1-Distill-Llama-70B wins long documents and following instructions; DeepSeek-R1-Distill-Qwen-32B wins coding and reasoning. DeepSeek-R1-Distill-Qwen-32B costs 68% less.
Which is better for coding?
DeepSeek-R1-Distill-Qwen-32B. It wins the one coding test both models report.
Which is cheaper?
DeepSeek-R1-Distill-Llama-70B costs $0.70 per million input tokens and $1.10 per million output tokens; DeepSeek-R1-Distill-Qwen-32B costs $0.29 and $0.29. That makes DeepSeek-R1-Distill-Qwen-32B about 68% cheaper for the same work.
How do you compare the two?
We use the 12 benchmark tests both models have published scores on. The verdict counts the 5 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 7 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.