DeepSeek-V2.5 vs Qwen2.5 Instruct 72B
Wins 0 of 3 areas
—
Wins 3 of 3 areas
Coding · Reasoning · Long documents
Qwen2.5 Instruct 72B is the stronger all-rounder.
Scores updated · 27 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.
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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Qwen2.5 Instruct 72B pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+7points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+6.8points ahead
- Questions about very long textsLongBench v2+4points ahead
Where DeepSeek-V2.5 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 27 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningQwen2.5 Instruct 72B
- GPQA DiamondQwen2.5 Instruct 72B by 6.842.349.1+6.8
Long documentsQwen2.5 Instruct 72B
- LongBench v2Qwen2.5 Instruct 72B by 435.439.4+4
Other results24 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.
- LiveCodeBenchQwen2.5 Instruct 72B by 13.741.855.5+13.7
- MATHQwen2.5 Instruct 72B by 13.774.788.4+13.7
- DROP (3-shot F1)DeepSeek-V2.5 by 11.187.876.7+11.1
- CodeforcesDeepSeek-V2.5 by 10.835.624.8+10.8
- CLUEWSCDeepSeek-V2.5 by 7.990.482.5+7.9
- AIME 2024Qwen2.5 Instruct 72B by 6.616.723.3+6.6
- MMLU-ReduxQwen2.5 Instruct 72B by 6.580.386.8+6.5
- Aider-Edit (Acc.)DeepSeek-V2.5 by 6.271.665.4+6.2
- Chinese SimpleQA (C-SimpleQA)DeepSeek-V2.5 by 5.754.148.4+5.7
- HellaSwagDeepSeek-V2.5 by 5.590.384.8+5.5
- MATH-500 (EM)Qwen2.5 Instruct 72B by 5.374.780+5.3
- Arena HardQwen2.5 Instruct 72B by 576.281.2+5
- MMLU-ProQwen2.5 Instruct 72B by 4.966.271.1+4.9
- BBHDeepSeek-V2.5 by 4.584.379.8+4.5
- FRAMES (Acc.)Qwen2.5 Instruct 72B by 4.465.469.8+4.4
- HumanEval-Mul (Pass@1)Qwen2.5 Instruct 72B by 3.573.877.3+3.5
- IFEvalQwen2.5 Instruct 72B by 3.580.684.1+3.5
- MMLUDeepSeek-V2.5 by 3.480.477+3.4
- HumanEvalDeepSeek-V2.5 by 2.48986.6+2.4
- AlignBenchQwen2.5 Instruct 72B by 1.280.481.6+1.2
- AA IntelligenceQwen2.5 Instruct 72B by 1.16.67.7+1.1
- SimpleQADeepSeek-V2.5 by 1.110.29.1+1.1
- GSM8Ktie95.195.8tie
- MT-Benchtie99.3tie
Questions people ask
Which is better, DeepSeek-V2.5 or Qwen2.5 Instruct 72B?
Qwen2.5 Instruct 72B wins all three areas where both have results: coding, reasoning and long documents. DeepSeek-V2.5 wins none.
Which is better for coding?
Qwen2.5 Instruct 72B. It wins the one coding test both models report.
How do you compare the two?
We use the 27 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 24 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.