Qwen3.5 35B A3B vs Qwen3.5 9B
Wins 6 of 8 areas
Coding · Reasoning · Images and charts · Math · Long documents · Following instructions
Wins 0 of 8 areas
—
Qwen3.5 35B A3B is the stronger all-rounder.Qwen3.5 9B is cheaper.
Scores updated · 41 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 software40Qwen3.5 35B A3B4 of 4 tests
- ReasoningHard problems that need careful thinking20Qwen3.5 35B A3B2 of 3 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video20Qwen3.5 35B A3B2 of 2 tests
- Long documentsFinding answers in very long texts20Qwen3.5 35B A3B2 of 2 tests
- Following instructionsDoing exactly what it is asked20Qwen3.5 35B A3B2 of 2 tests
- MathCompetition and research-level math10Qwen3.5 35B A3B1 of 2 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own11Even1 each
- FactsGetting facts right instead of making them up11Even1 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.
Qwen3.5 9B costs 85% 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 Qwen3.5 35B A3B pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+11.6points ahead
- Harvard-MIT high-school math contest problemsHMMT Feb. 2026+10.6points ahead
- Recent programming contest problemsLiveCodeBench v6+9points ahead
Where Qwen3.5 9B pulls ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+2.1points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+1.8points ahead
Every test, side by side
All 41 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3.5 35B A3B
- Terminal-Bench 2.1Qwen3.5 35B A3B by 11.640.829.2+11.6
- LiveCodeBench v6Qwen3.5 35B A3B by 974.665.6+9
- SciCodeQwen3.5 35B A3B by 8.237.729.5+8.2
- Terminal-Bench HardQwen3.5 35B A3B by 2.326.524.2+2.3
AgentsEven
- GDPValQwen3.5 35B A3B by 5.65.60+5.6
- τ-Bench V3 · BankingQwen3.5 9B by 2.14.97+2.1
ReasoningQwen3.5 35B A3B
- Humanity's Last ExamQwen3.5 35B A3B by 6.12114.9+6.1
- GPQA DiamondQwen3.5 35B A3B by 3.984.580.6+3.9
- CritPttie0.90.3tie
FactsEven
- AA-Omniscience · AccuracyQwen3.5 35B A3B by 3.720.116.4+3.7
- AA-Omniscience · Non-hallucinationQwen3.5 9B by 1.814.616.4+1.8
Images and chartsQwen3.5 35B A3B
- OCRBenchv2Qwen3.5 35B A3B by 6.665.358.7+6.6
- MMMU-ProQwen3.5 35B A3B by 3.572.769.2+3.5
MathQwen3.5 35B A3B
- HMMT Feb. 2026Qwen3.5 35B A3B by 10.681.871.2+10.6
- AIME 2026tie93.392.5tie
Long documentsQwen3.5 35B A3B
- LongBench v2Qwen3.5 35B A3B by 3.85955.2+3.8
- AA-LCRQwen3.5 35B A3B by 27270+2
Following instructionsQwen3.5 35B A3B
- IFBenchQwen3.5 35B A3B by 5.872.566.7+5.8
- Multi-ChallengeQwen3.5 35B A3B by 5.56054.5+5.5
Other results22 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.
- MLVUQwen3.5 35B A3B by 33.285.652.4+33.2
- Artificial Analysis Coding IndexQwen3.5 35B A3B by 8.33728.7+8.3
- AA IntelligenceQwen3.5 35B A3B by 8.119.311.2+8.1
- BFCLv4Qwen3.5 35B A3B by 7.267.360.1+7.2
- HMMT 2025Qwen3.5 35B A3B by 6.389.282.9+6.3
- AA-OmniscienceQwen3.5 35B A3B by 5.4-48.1-53.5+5.4
- SuperGPQAQwen3.5 35B A3B by 5.263.458.2+5.2
- AA Agentic IndexQwen3.5 35B A3B by 4.811.87+4.8
- DeepPlanningQwen3.5 35B A3B by 4.822.818+4.8
- MMLU-ProXQwen3.5 35B A3B by 4.78176.3+4.7
- IncludeQwen3.5 35B A3B by 4.179.775.6+4.1
- MMMLUQwen3.5 35B A3B by 485.281.2+4
- MMVUQwen3.5 35B A3B by 3.772.368.6+3.7
- MMLU-ProQwen3.5 35B A3B by 2.885.382.5+2.8
- OCRBench v2 (Chinese)Qwen3.5 9B by 2.461.764.1+2.4
- τ²-Bench Telecom (AA run)Qwen3.5 35B A3B by 2.489.286.8+2.4
- MMLU-ReduxQwen3.5 35B A3B by 2.293.391.1+2.2
- t2-benchQwen3.5 35B A3B by 2.181.279.1+2.1
- VITA-BenchQwen3.5 35B A3B by 2.131.929.8+2.1
- C-EvalQwen3.5 35B A3B by 290.288.2+2
- Claw-Eval AvgQwen3.5 9B by 1.165.466.5+1.1
- IFEvaltie91.991.5tie
Questions people ask
Which is better, Qwen3.5 35B A3B or Qwen3.5 9B?
Qwen3.5 35B A3B wins six of the eight areas we test: coding, reasoning, images and charts, math, long documents and following instructions. Qwen3.5 9B wins none, but costs 85% less. They are level on agents and facts.
Which is better for coding?
Qwen3.5 35B A3B. It wins 4 of the 4 coding tests both models report; Qwen3.5 9B wins none.
Which is cheaper?
Qwen3.5 35B A3B costs $0.25 per million input tokens and $2.00 per million output tokens; Qwen3.5 9B costs $0.14 and $0.20. That makes Qwen3.5 9B about 85% cheaper for the same work.
How do you compare the two?
We use the 41 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 22 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.