Claude Opus 4.5 vs Qwen3.5 35B A3B
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Math · Long documents
Wins 2 of 8 areas
Images and charts · Following instructions
Claude Opus 4.5 is the stronger all-rounder.Qwen3.5 35B A3B is cheaper and better at images and charts.
Scores updated · 57 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.
- CodingWriting and fixing software61Claude Opus 4.56 of 7 tests
- ReasoningHard problems that need careful thinking30Claude Opus 4.53 of 3 tests
- MathCompetition and research-level math30Claude Opus 4.53 of 3 tests
- FactsGetting facts right instead of making them up20Claude Opus 4.52 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts20Claude Opus 4.52 of 2 tests
- AgentsCarrying out multi-step tasks on its own21Claude Opus 4.52 of 4 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video13Qwen3.5 35B A3B3 of 5 tests · 1 tie
- Following instructionsDoing exactly what it is asked02Qwen3.5 35B A3B2 of 2 tests
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 35B A3B costs 93% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Opus 4.5 pulls ahead
- Real work tasks from 44 professionsGDPVal+41.7points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+26.5points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+24.4points ahead
Where Qwen3.5 35B A3B pulls ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+53.3points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+24points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+14.5points ahead
Every test, side by side
All 57 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.5
- SWE-bench MultilingualQwen3.5 35B A3B by 53.3760.3+53.3
- LMArena · WebDevClaude Opus 4.5 by 218 rating points14681250+218 rating
- Terminal-Bench HardClaude Opus 4.5 by 20.54726.5+20.5
- SciCodeClaude Opus 4.5 by 11.849.537.7+11.8
- SWE-bench VerifiedClaude Opus 4.5 by 11.780.969.2+11.7
- LiveCodeBench v6Claude Opus 4.5 by 10.284.874.6+10.2
- SWE-bench ProClaude Opus 4.5 by 7.45244.6+7.4
AgentsClaude Opus 4.5
- GDPValClaude Opus 4.5 by 41.747.35.6+41.7
- BrowseCompQwen3.5 35B A3B by 243761+24
- OSWorld-VerifiedClaude Opus 4.5 by 11.866.354.5+11.8
- MCP Atlastie62.362.4tie
ReasoningClaude Opus 4.5
- Humanity's Last ExamClaude Opus 4.5 by 9.130.121+9.1
- CritPtClaude Opus 4.5 by 3.74.60.9+3.7
- GPQA DiamondClaude Opus 4.5 by 2.186.684.5+2.1
FactsClaude Opus 4.5
- AA-Omniscience · AccuracyClaude Opus 4.5 by 26.546.620.1+26.5
- AA-Omniscience · Non-hallucinationClaude Opus 4.5 by 24.43914.6+24.4
- Vectara HHEM hallucination ratelower is bettertie10.910.5tie
Images and chartsQwen3.5 35B A3B
- CharXiv (RQ)Qwen3.5 35B A3B by 10.367.277.5+10.3
- MathVistaQwen3.5 35B A3B by 680.286.2+6
- OCRBenchQwen3.5 35B A3B by 4.586.591+4.5
- MMMU-ProClaude Opus 4.5 by 1.37472.7+1.3
- MMMUtie80.781.4tie
MathClaude Opus 4.5
- IMOAnswerBenchClaude Opus 4.5 by 7.28476.8+7.2
- HMMT Feb. 2026Claude Opus 4.5 by 3.585.381.8+3.5
- AIME 2026Claude Opus 4.5 by 1.895.193.3+1.8
Long documentsClaude Opus 4.5
- LongBench v2Claude Opus 4.5 by 5.464.459+5.4
- AA-LCRClaude Opus 4.5 by 5.377.372+5.3
Following instructionsQwen3.5 35B A3B
- IFBenchQwen3.5 35B A3B by 14.55872.5+14.5
- Multi-ChallengeQwen3.5 35B A3B by 15960+1
Other results28 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 Opus 4.5 by 62.114-48.1+62.1
- QwenWebBenchClaude Opus 4.5 by 558 rating points1536978+558 rating
- AA Agentic IndexClaude Opus 4.5 by 47.859.611.8+47.8
- SkillsBench Avg5Claude Opus 4.5 by 40.945.34.4+40.9
- NL2RepoClaude Opus 4.5 by 22.743.220.5+22.7
- Terminal-Bench 2.0Claude Opus 4.5 by 18.859.340.5+18.8
- SimpleVQAClaude Opus 4.5 by 11.469.758.3+11.4
- Claw-Eval AvgClaude Opus 4.5 by 11.276.665.4+11.2
- Artificial Analysis Coding IndexClaude Opus 4.5 by 10.847.837+10.8
- AA IntelligenceClaude Opus 4.5 by 9.829.119.3+9.8
- SuperGPQAClaude Opus 4.5 by 7.270.663.4+7.2
- MATH-VisionQwen3.5 35B A3B by 6.877.183.9+6.8
- Seal-0Claude Opus 4.5 by 6.347.741.4+6.3
- MMMLUClaude Opus 4.5 by 5.690.885.2+5.6
- MMVUClaude Opus 4.5 by 577.372.3+5
- ZeroBenchQwen3.5 35B A3B by 538+5
- QwenClawBenchClaude Opus 4.5 by 4.652.347.7+4.6
- MMLU-ProClaude Opus 4.5 by 4.289.585.3+4.2
- HMMT Nov. 2025Claude Opus 4.5 by 4.193.389.2+4.1
- VideoMMMUClaude Opus 4.5 by 484.480.4+4
- HMMT Feb. 2025Claude Opus 4.5 by 3.992.989+3.9
- MMLU-ReduxClaude Opus 4.5 by 2.395.693.3+2.3
- C-EvalClaude Opus 4.5 by 292.290.2+2
- vectara_avg_summary_lengthClaude Opus 4.5 by 19.6 rating points114.594.9+19.6 rating
- OmniDocBench 1.5Qwen3.5 35B A3B by 1.687.789.3+1.6
- vectara_answer_rateQwen3.5 35B A3B by 1.198.799.8+1.1
- vectara_factual_consistencytie89.189.5tie
- τ²-Bench Telecom (AA run)tie89.589.2tie
Questions people ask
Which is better, Claude Opus 4.5 or Qwen3.5 35B A3B?
Claude Opus 4.5 wins six of the eight areas we test: coding, agents, reasoning, facts, math and long documents. Qwen3.5 35B A3B wins images and charts and following instructions, and costs 93% less.
Which is better for coding?
Claude Opus 4.5. It wins 6 of the 7 coding tests both models report; Qwen3.5 35B A3B wins 1.
Which is cheaper?
Claude Opus 4.5 costs $5.00 per million input tokens and $25.00 per million output tokens; Qwen3.5 35B A3B costs $0.25 and $2.00. That makes Qwen3.5 35B A3B about 93% cheaper for the same work.
How do you compare the two?
We use the 57 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 28 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.