Qwen3.5 4B vs Qwen3.5 9B
Wins 0 of 8 areas
—
Wins 6 of 8 areas
Coding · Reasoning · Facts · Images and charts · Long documents · Following instructions
Qwen3.5 9B is the stronger all-rounder.Qwen3.5 4B 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 software04Qwen3.5 9B4 of 4 tests
- ReasoningHard problems that need careful thinking02Qwen3.5 9B2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up02Qwen3.5 9B2 of 2 tests
- Long documentsFinding answers in very long texts02Qwen3.5 9B2 of 2 tests
- Following instructionsDoing exactly what it is asked02Qwen3.5 9B2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video01Qwen3.5 9B1 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- MathCompetition and research-level math11Even1 each
Agents and images and charts 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.
Qwen3.5 4B costs 46% 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 9B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+14.7points ahead
- Code for real scientific research problemsSciCode+13.4points ahead
- Recent programming contest problemsLiveCodeBench v6+9.8points ahead
Where Qwen3.5 4B pulls ahead
- Harvard-MIT high-school math contest problemsHMMT Feb. 2026+1.7points 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 9B
- SciCodeQwen3.5 9B by 13.416.129.5+13.4
- LiveCodeBench v6Qwen3.5 9B by 9.855.865.6+9.8
- Terminal-Bench HardQwen3.5 9B by 618.224.2+6
- Terminal-Bench 2.1Qwen3.5 9B by 3.425.829.2+3.4
ReasoningQwen3.5 9B
- Humanity's Last ExamQwen3.5 9B by 59.914.9+5
- GPQA DiamondQwen3.5 9B by 3.577.180.6+3.5
- CritPttie00.3tie
FactsQwen3.5 9B
- AA-Omniscience · Non-hallucinationQwen3.5 9B by 313.416.4+3
- AA-Omniscience · AccuracyQwen3.5 9B by 1.315.116.4+1.3
MathEven
- AIME 2026Qwen3.5 9B by 2.889.792.5+2.8
- HMMT Feb. 2026Qwen3.5 4B by 1.772.971.2+1.7
Long documentsQwen3.5 9B
- AA-LCRQwen3.5 9B by 76370+7
- LongBench v2Qwen3.5 9B by 5.25055.2+5.2
Following instructionsQwen3.5 9B
- IFBenchQwen3.5 9B by 14.75266.7+14.7
- Multi-ChallengeQwen3.5 9B by 5.54954.5+5.5
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.
- AA Agentic IndexQwen3.5 4B by 25.532.57+25.5
- BFCLv4Qwen3.5 9B by 15.850.366.1+15.8
- LiveCodeBenchV6 no toolsQwen3.5 9B by 960.969.9+9
- VITA-BenchQwen3.5 9B by 7.82229.8+7.8
- AIME25 no toolsQwen3.5 9B by 6.849.356.1+6.8
- Artificial Analysis Coding IndexQwen3.5 9B by 6.122.628.7+6.1
- HMMT 2025Qwen3.5 9B by 6.176.882.9+6.1
- IFEvalQwen3.5 9B by 5.386.291.5+5.3
- SuperGPQAQwen3.5 9B by 5.352.958.2+5.3
- τ²-Bench Telecom (AA run)Qwen3.5 4B by 5.392.186.8+5.3
- MMMLUQwen3.5 9B by 5.176.181.2+5.1
- AA-OmniscienceQwen3.5 9B by 4.8-58.4-53.5+4.8
- MMLU-ProXQwen3.5 9B by 4.871.576.3+4.8
- IncludeQwen3.5 9B by 4.67175.6+4.6
- Claw-Eval AvgQwen3.5 9B by 4.262.366.5+4.2
- MMLU-ProQwen3.5 9B by 3.479.182.5+3.4
- C-EvalQwen3.5 9B by 3.185.188.2+3.1
- BrowseComp+ (OpenClaw)Qwen3.5 9B by 2.724.527.2+2.7
- MMLU-ReduxQwen3.5 9B by 2.388.891.1+2.3
- AA IntelligenceQwen3.5 4B by 1.913.111.2+1.9
- t2-benchtie79.979.1tie
- DeepPlanningtie17.618tie
- τ³-Bench Bankingtie5.55.2tie
- PinchBenchtie71.371.5tie
Questions people ask
Which is better, Qwen3.5 4B or Qwen3.5 9B?
Qwen3.5 9B wins six of the eight areas we test: coding, reasoning, facts, images and charts, long documents and following instructions. Qwen3.5 4B wins none, but costs 46% less. They are level on agents and math.
Which is better for coding?
Qwen3.5 9B. It wins 4 of the 4 coding tests both models report; Qwen3.5 4B wins none.
Which is cheaper?
Qwen3.5 4B costs $0.03 per million input tokens and $0.15 per million output tokens; Qwen3.5 9B costs $0.14 and $0.20. That makes Qwen3.5 4B about 46% 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 17 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.