Qwen3 VL 32B Reasoning vs Qwen3 VL 4B Instruct
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
Wins 0 of 7 areas
—
Qwen3 VL 32B Reasoning is the stronger all-rounder.
Scores updated · 49 tests both models report · How we compare
Where each one wins
Tests won in each of the seven 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.
- Images and chartsUnderstanding pictures, charts and video51Qwen3 VL 32B Reasoning5 of 6 tests
- CodingWriting and fixing software30Qwen3 VL 32B Reasoning3 of 3 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 32B Reasoning2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own20Qwen3 VL 32B Reasoning2 of 2 tests
- FactsGetting facts right instead of making them up20Qwen3 VL 32B Reasoning2 of 2 tests
- Long documentsFinding answers in very long texts10Qwen3 VL 32B Reasoning1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 VL 32B Reasoning1 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.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Qwen3 VL 32B Reasoning pulls ahead
- Reasons across sets of long documentsAA-LCR+41.3points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+36.2points ahead
- Recent programming contest problemsLiveCodeBench v6+27.7points ahead
Where Qwen3 VL 4B Instruct pulls ahead
- Reads text in imagesOCRBench+2.6points 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.
CodingQwen3 VL 32B Reasoning
- LiveCodeBench v6Qwen3 VL 32B Reasoning by 27.765.637.9+27.7
- SciCodeQwen3 VL 32B Reasoning by 14.828.513.7+14.8
- Terminal-Bench HardQwen3 VL 32B Reasoning by 7.67.60+7.6
AgentsQwen3 VL 32B Reasoning
- OSWorld-VerifiedQwen3 VL 32B Reasoning by 14.84126.2+14.8
- GDPValQwen3 VL 32B Reasoning by 7.37.30+7.3
ReasoningQwen3 VL 32B Reasoning
- GPQA DiamondQwen3 VL 32B Reasoning by 36.273.337.1+36.2
- Humanity's Last ExamQwen3 VL 32B Reasoning by 6.510.13.6+6.5
- CritPttie00tie
FactsQwen3 VL 32B Reasoning
- AA-Omniscience · Non-hallucinationQwen3 VL 32B Reasoning by 13.716.32.6+13.7
- AA-Omniscience · AccuracyQwen3 VL 32B Reasoning by 6.11710.9+6.1
Images and chartsQwen3 VL 32B Reasoning
Long documentsQwen3 VL 32B Reasoning
- AA-LCRQwen3 VL 32B Reasoning by 41.355.314+41.3
Following instructionsQwen3 VL 32B Reasoning
- IFBenchQwen3 VL 32B Reasoning by 27.659.431.8+27.6
Other results31 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.
- AIME 2025Qwen3 VL 32B Reasoning by 37.183.746.6+37.1
- AA-OmniscienceQwen3 VL 32B Reasoning by 23.5-52.5-76+23.5
- VideoMMMUQwen3 VL 32B Reasoning by 22.87956.2+22.8
- τ²-Bench Telecom (AA run)Qwen3 VL 32B Reasoning by 22.245.623.4+22.2
- SuperGPQAQwen3 VL 32B Reasoning by 18.75940.3+18.7
- MATH-VisionQwen3 VL 32B Reasoning by 18.670.251.6+18.6
- MMLU-ProXQwen3 VL 32B Reasoning by 17.877.259.4+17.8
- AA Agentic IndexQwen3 VL 32B Reasoning by 15.623.47.8+15.6
- MMLU-ProQwen3 VL 32B Reasoning by 1582.167.1+15
- IncludeQwen3 VL 32B Reasoning by 14.976.361.4+14.9
- ERQAQwen3 VL 32B Reasoning by 1152.341.3+11
- MMMU (val) (Pass@1)Qwen3 VL 32B Reasoning by 10.778.167.4+10.7
- MMLU-ReduxQwen3 VL 32B Reasoning by 10.491.981.5+10.4
- Artificial Analysis Coding IndexQwen3 VL 32B Reasoning by 9.914.54.6+9.9
- HallusionBenchQwen3 VL 32B Reasoning by 9.867.457.6+9.8
- InfoVQA_testQwen3 VL 32B Reasoning by 8.989.280.3+8.9
- BFCL v3Qwen3 VL 32B Reasoning by 8.471.763.3+8.4
- RealWorldQAQwen3 VL 32B Reasoning by 7.578.470.9+7.5
- SimpleQAQwen3 VL 32B Reasoning by 7.455.448+7.4
- CharadesSTAQwen3 VL 32B Reasoning by 7.362.855.5+7.3
- LVBenchQwen3 VL 32B Reasoning by 6.462.656.2+6.4
- AA IntelligenceQwen3 VL 32B Reasoning by 6.211.95.7+6.2
- IFEvalQwen3 VL 32B Reasoning by 5.587.882.3+5.5
- AI2DQwen3 VL 32B Reasoning by 4.888.984.1+4.8
- OCRBench v2_enQwen3 VL 32B Reasoning by 4.768.463.7+4.7
- OCRBench v2 (Chinese)Qwen3 VL 32B Reasoning by 4.562.157.6+4.5
- MV-BenchQwen3 VL 32B Reasoning by 4.373.268.9+4.3
- ScreenSpot-Pro (No tools)Qwen3 VL 4B Instruct by 2.457.159.5+2.4
- ScreenSpotQwen3 VL 32B Reasoning by 1.795.794+1.7
- DocVQA_testtie96.195.3tie
- MM MTBenchtie8.37.5tie
Questions people ask
Which is better, Qwen3 VL 32B Reasoning or Qwen3 VL 4B Instruct?
Qwen3 VL 32B Reasoning wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, long documents and following instructions. Qwen3 VL 4B Instruct wins none.
Which is better for coding?
Qwen3 VL 32B Reasoning. It wins 3 of the 3 coding tests both models report; Qwen3 VL 4B Instruct wins none.
How do you compare the two?
We use the 49 benchmark tests both models have published scores on. The verdict counts the 18 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 31 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.