Claude Sonnet 4.5 vs Qwen3.5 27B
Wins 2 of 8 areas
Coding · Facts
Wins 5 of 8 areas
Agents · Reasoning · Images and charts · Math · Following instructions
Qwen3.5 27B is the stronger all-rounder.Claude Sonnet 4.5 is better at facts.
Scores updated · 49 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- Images and chartsUnderstanding pictures, charts and video04Qwen3.5 27B4 of 4 tests
- AgentsCarrying out multi-step tasks on its own13Qwen3.5 27B3 of 4 tests
- ReasoningHard problems that need careful thinking02Qwen3.5 27B2 of 3 tests · 1 tie
- Following instructionsDoing exactly what it is asked02Qwen3.5 27B2 of 2 tests
- MathCompetition and research-level math01Qwen3.5 27B1 of 1 test
- FactsGetting facts right instead of making them up20Claude Sonnet 4.52 of 3 tests · 1 tie
- CodingWriting and fixing software43Claude Sonnet 4.54 of 7 tests
- Long documentsFinding answers in very long texts11Even1 each
Math rests on a single test.
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.
Qwen3.5 27B costs 85% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3.5 27B pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+36.9points ahead
- Multi-step tasks using many real software toolsMCP Atlas+24.6points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+18.3points ahead
Where Claude Sonnet 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+32.4points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+12.2points ahead
- Code for real scientific research problemsSciCode+6.2points ahead
Every test, side by side
All 49 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4.5
- LiveCodeBench v6Qwen3.5 27B by 16.76480.7+16.7
- SWE-bench ProQwen3.5 27B by 7.643.651.2+7.6
- SciCodeClaude Sonnet 4.5 by 6.245.739.5+6.2
- SWE-bench VerifiedClaude Sonnet 4.5 by 4.877.272.4+4.8
- LMArena · WebDevClaude Sonnet 4.5 by 36 rating points13931357+36 rating
- Terminal-Bench HardClaude Sonnet 4.5 by 335.632.6+3
- SWE-bench MultilingualQwen3.5 27B by 2.36769.3+2.3
AgentsQwen3.5 27B
- BrowseCompQwen3.5 27B by 36.924.161+36.9
- MCP AtlasQwen3.5 27B by 24.643.868.4+24.6
- GDPValQwen3.5 27B by 12.520.533+12.5
- OSWorld-VerifiedClaude Sonnet 4.5 by 5.261.456.2+5.2
ReasoningQwen3.5 27B
- Humanity's Last ExamQwen3.5 27B by 6.117.823.9+6.1
- GPQA DiamondQwen3.5 27B by 2.483.485.8+2.4
- CritPttie1.10.9tie
FactsClaude Sonnet 4.5
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.5 by 32.450.918.5+32.4
- AA-Omniscience · AccuracyClaude Sonnet 4.5 by 12.232.920.7+12.2
- Vectara HHEM hallucination ratelower is bettertie1212.1tie
Images and chartsQwen3.5 27B
- CharXiv (RQ)Qwen3.5 27B by 12.367.279.5+12.3
- MathVistaQwen3.5 27B by 879.887.8+8
- MMMU-ProQwen3.5 27B by 6.368.775+6.3
- MMMUQwen3.5 27B by 2.779.682.3+2.7
Long documentsEven
- AA-LCRQwen3.5 27B by 5.472.377.7+5.4
- LongBench v2Claude Sonnet 4.5 by 1.261.860.6+1.2
Following instructionsQwen3.5 27B
- IFBenchQwen3.5 27B by 18.357.375.6+18.3
- Multi-ChallengeQwen3.5 27B by 5.555.360.8+5.5
Other results23 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 4.5 by 43.9-0.1-44+43.9
- AA Agentic IndexQwen3.5 27B by 37.117.554.6+37.1
- HMMT Feb. 2025Qwen3.5 27B by 24.567.592+24.5
- BrowseComp-ZHQwen3.5 27B by 19.742.462.1+19.7
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 17.252.134.9+17.2
- τ²-Bench Telecom (AA run)Qwen3.5 27B by 15.878.193.9+15.8
- HMMT 2025Qwen3.5 27B by 15.274.689.8+15.2
- RealWorldQAQwen3.5 27B by 13.470.383.7+13.4
- CC-OCRQwen3.5 27B by 12.968.181+12.9
- HallusionBenchQwen3.5 27B by 10.159.970+10.1
- Terminal-Bench 2.0Claude Sonnet 4.5 by 8.45041.6+8.4
- HMMT Nov. 2025Qwen3.5 27B by 8.181.789.8+8.1
- AI2DQwen3.5 27B by 5.98792.9+5.9
- MMBench EN-DEV-v1.1Qwen3.5 27B by 4.388.392.6+4.3
- vectara_answer_rateQwen3.5 27B by 4.295.699.8+4.2
- vectara_avg_summary_lengthClaude Sonnet 4.5 by 33.4 rating points127.894.4+33.4 rating
- MMMLUClaude Sonnet 4.5 by 3.289.185.9+3.2
- OmniDocBench 1.5Qwen3.5 27B by 3.185.888.9+3.1
- MMLU-ReduxClaude Sonnet 4.5 by 2.495.693.2+2.4
- AA IntelligenceQwen3.5 27B by 2.220.722.9+2.2
- SimpleVQAClaude Sonnet 4.5 by 1.657.656+1.6
- MMLU-ProClaude Sonnet 4.5 by 1.487.586.1+1.4
- vectara_factual_consistencytie8887.9tie
Questions people ask
Which is better, Claude Sonnet 4.5 or Qwen3.5 27B?
Qwen3.5 27B wins five of the eight areas we test: agents, reasoning, images and charts, math and following instructions. Claude Sonnet 4.5 wins coding and facts. They are level on long documents.
Which is better for coding?
Claude Sonnet 4.5. It wins 4 of the 7 coding tests both models report; Qwen3.5 27B wins 3.
Which is cheaper?
Claude Sonnet 4.5 costs $3.00 per million input tokens and $15.00 per million output tokens; Qwen3.5 27B costs $0.30 and $2.40. That makes Qwen3.5 27B about 85% cheaper for the same work.
How do you compare the two?
We use the 49 benchmark tests both models have published scores on. The verdict counts the 26 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 23 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.