Qwen3 VL 32B Instruct vs Qwen3 VL 4B (Reasoning)
Wins 6 of 7 areas
Coding · Reasoning · Facts · Images and charts · Long documents · Following instructions
Wins 0 of 7 areas
—
Qwen3 VL 32B Instruct 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 video60Qwen3 VL 32B Instruct6 of 6 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 32B Instruct2 of 3 tests · 1 tie
- CodingWriting and fixing software21Qwen3 VL 32B Instruct2 of 3 tests
- FactsGetting facts right instead of making them up10Qwen3 VL 32B Instruct1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10Qwen3 VL 32B Instruct1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 VL 32B Instruct1 of 1 test
- AgentsCarrying out multi-step tasks on its own11Even1 each
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 Instruct pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+17.7points ahead
- Code for real scientific research problemsSciCode+13points ahead
- Reasons across sets of long documentsAA-LCR+13points ahead
Where Qwen3 VL 4B (Reasoning) pulls ahead
- Recent programming contest problemsLiveCodeBench v6+7.5points 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 Instruct
- SciCodeQwen3 VL 32B Instruct by 1330.117.1+13
- LiveCodeBench v6Qwen3 VL 4B (Reasoning) by 7.543.851.3+7.5
- Terminal-Bench HardQwen3 VL 32B Instruct by 6.88.31.5+6.8
AgentsEven
- GDPValQwen3 VL 4B (Reasoning) by 13.8013.8+13.8
- OSWorld-VerifiedQwen3 VL 32B Instruct by 1.232.631.4+1.2
ReasoningQwen3 VL 32B Instruct
- GPQA DiamondQwen3 VL 32B Instruct by 17.767.149.4+17.7
- Humanity's Last ExamQwen3 VL 32B Instruct by 2.26.84.6+2.2
- CritPttie00tie
FactsQwen3 VL 32B Instruct
- AA-Omniscience · AccuracyQwen3 VL 32B Instruct by 2.614.812.2+2.6
- AA-Omniscience · Non-hallucinationtie8.28.2tie
Images and chartsQwen3 VL 32B Instruct
Long documentsQwen3 VL 32B Instruct
- AA-LCRQwen3 VL 32B Instruct by 1334.321.3+13
Following instructionsQwen3 VL 32B Instruct
- IFBenchQwen3 VL 32B Instruct by 2.639.236.6+2.6
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.
- Arena-Hard-V2Qwen3 VL 32B Instruct by 27.964.736.8+27.9
- τ²-Bench Telecom (AA run)Qwen3 VL 32B Instruct by 13.729.215.5+13.7
- LVBenchQwen3 VL 32B Instruct by 10.363.853.5+10.3
- IncludeQwen3 VL 32B Instruct by 9.47464.6+9.4
- Artificial Analysis Coding IndexQwen3 VL 32B Instruct by 8.915.66.7+8.9
- ScreenSpot-Pro (No tools)Qwen3 VL 32B Instruct by 8.757.949.2+8.7
- MMLU-ProXQwen3 VL 32B Instruct by 8.473.465+8.4
- AIME 2025Qwen3 VL 4B (Reasoning) by 8.366.274.5+8.3
- SuperGPQAQwen3 VL 32B Instruct by 7.854.646.8+7.8
- CC-OCRQwen3 VL 32B Instruct by 6.580.373.8+6.5
- RealWorldQAQwen3 VL 32B Instruct by 5.87973.2+5.8
- OCRBench v2_enQwen3 VL 32B Instruct by 5.667.461.8+5.6
- MMMU (val) (Pass@1)Qwen3 VL 32B Instruct by 5.27670.8+5.2
- AA-OmniscienceQwen3 VL 32B Instruct by 5-63.4-68.4+5
- MMLU-ProQwen3 VL 32B Instruct by 578.673.6+5
- AA Agentic IndexQwen3 VL 4B (Reasoning) by 4.79.714.4+4.7
- AI2DQwen3 VL 32B Instruct by 4.689.584.9+4.6
- InfoVQA_testQwen3 VL 32B Instruct by 48783+4
- MMLU-ReduxQwen3 VL 32B Instruct by 3.889.886+3.8
- MV-BenchQwen3 VL 32B Instruct by 3.572.869.3+3.5
- MATH-VisionQwen3 VL 32B Instruct by 3.463.460+3.4
- OCRBench v2 (Chinese)Qwen3 VL 32B Instruct by 3.459.255.8+3.4
- BFCL v3Qwen3 VL 32B Instruct by 2.970.267.3+2.9
- ScreenSpotQwen3 VL 32B Instruct by 2.995.892.9+2.9
- DocVQA_testQwen3 VL 32B Instruct by 2.796.994.2+2.7
- CharadesSTAQwen3 VL 32B Instruct by 2.261.259+2.2
- IFEvalQwen3 VL 32B Instruct by 2.184.782.6+2.1
- ERQAQwen3 VL 32B Instruct by 1.548.847.3+1.5
- AA IntelligenceQwen3 VL 32B Instruct by 1.48.47+1.4
- MM MTBenchtie8.47.7tie
- HallusionBenchtie63.864.1tie
Questions people ask
Which is better, Qwen3 VL 32B Instruct or Qwen3 VL 4B (Reasoning)?
Qwen3 VL 32B Instruct wins six of the seven areas where both have results: coding, reasoning, facts, images and charts, long documents and following instructions. Qwen3 VL 4B (Reasoning) wins none. They are level on agents.
Which is better for coding?
Qwen3 VL 32B Instruct. It wins 2 of the 3 coding tests both models report; Qwen3 VL 4B (Reasoning) wins 1.
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.