Claude Sonnet 4 vs Qwen2.5 Instruct 72B
Wins 5 of 5 areas
Coding · Reasoning · Facts · Long documents · Following instructions
Wins 0 of 5 areas
—
Claude Sonnet 4 is the stronger all-rounder.Qwen2.5 Instruct 72B is cheaper.
Scores updated · 31 tests both models report · How we compare
Where each one wins
Tests won in each of the five 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.
- CodingWriting and fixing software30Claude Sonnet 43 of 3 tests
- ReasoningHard problems that need careful thinking20Claude Sonnet 42 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up20Claude Sonnet 42 of 2 tests
- Long documentsFinding answers in very long texts10Claude Sonnet 41 of 1 test
- Following instructionsDoing exactly what it is asked10Claude Sonnet 41 of 1 test
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.
Qwen2.5 Instruct 72B costs 95% 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 Claude Sonnet 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+56.4points ahead
- Reasons across sets of long documentsAA-LCR+50.3points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+48.9points ahead
Where Qwen2.5 Instruct 72B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 31 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4
- SWE-bench VerifiedClaude Sonnet 4 by 48.972.723.8+48.9
- Terminal-Bench HardClaude Sonnet 4 by 26.631.14.5+26.6
- SciCodeClaude Sonnet 4 by 13.34026.7+13.3
ReasoningClaude Sonnet 4
- GPQA DiamondClaude Sonnet 4 by 28.677.749.1+28.6
- Humanity's Last ExamClaude Sonnet 4 by 7.110.73.6+7.1
- CritPttie0.30tie
FactsClaude Sonnet 4
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 56.470.914.5+56.4
- AA-Omniscience · AccuracyClaude Sonnet 4 by 5.222.717.5+5.2
Following instructionsClaude Sonnet 4
- IFBenchClaude Sonnet 4 by 17.854.736.9+17.8
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.
- AA-OmniscienceClaude Sonnet 4 by 53.30.2-53.1+53.3
- τ²-Bench Telecom (AA run)Claude Sonnet 4 by 30.164.634.5+30.1
- AIR-Bench 2024Claude Sonnet 4 by 29.388.359+29.3
- Artificial Analysis Coding IndexClaude Sonnet 4 by 25.737.611.9+25.7
- HarmBenchClaude Sonnet 4 by 25.29872.8+25.2
- LiveBenchClaude Sonnet 4 by 22.574.852.3+22.5
- AIME 2024Claude Sonnet 4 by 20.143.423.3+20.1
- MMLUClaude Sonnet 4 by 14.591.577+14.5
- MATH-500 (EM)Claude Sonnet 4 by 149480+14
- MultiPL-EClaude Sonnet 4 by 13.588.675.1+13.5
- LiveCodeBenchClaude Sonnet 4 by 1368.555.5+13
- MMLU-ProClaude Sonnet 4 by 12.683.771.1+12.6
- MMMLUClaude Sonnet 4 by 11.786.574.8+11.7
- AA IntelligenceClaude Sonnet 4 by 11.218.97.7+11.2
- MMLU-ReduxClaude Sonnet 4 by 6.893.686.8+6.8
- SimpleQAClaude Sonnet 4 by 6.815.99.1+6.8
- IFEvalClaude Sonnet 4 by 3.587.684.1+3.5
- XSTestQwen2.5 Instruct 72B by 1.396.697.9+1.3
- bbqClaude Sonnet 4 by 1.296.695.4+1.2
- anthropic_red_teamtie99.299.6tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, Claude Sonnet 4 or Qwen2.5 Instruct 72B?
Claude Sonnet 4 wins all five areas where both have results: coding, reasoning, facts, long documents and following instructions. Qwen2.5 Instruct 72B wins none, but costs 95% less.
Which is better for coding?
Claude Sonnet 4. It wins 3 of the 3 coding tests both models report; Qwen2.5 Instruct 72B wins none.
Which is cheaper?
Claude Sonnet 4 costs $3.00 per million input tokens and $15.00 per million output tokens; Qwen2.5 Instruct 72B costs $0.47 and $0.49. That makes Qwen2.5 Instruct 72B about 95% cheaper for the same work.
How do you compare the two?
We use the 31 benchmark tests both models have published scores on. The verdict counts the 10 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.