Qwen3 VL 32B Instruct vs Qwen3 VL 8B Instruct
Wins 5 of 7 areas
Coding · Reasoning · Images and charts · Long documents · Following instructions
Wins 2 of 7 areas
Agents · Facts
Qwen3 VL 32B Instruct is the stronger all-rounder.Qwen3 VL 8B Instruct is better at agents.
Scores updated · 48 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 video41Qwen3 VL 32B Instruct4 of 6 tests · 1 tie
- CodingWriting and fixing software30Qwen3 VL 32B Instruct3 of 3 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 32B Instruct2 of 3 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 own02Qwen3 VL 8B Instruct2 of 2 tests
- FactsGetting facts right instead of making them up01Qwen3 VL 8B Instruct1 of 2 tests · 1 tie
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.
Qwen3 VL 32B Instruct costs 9% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Qwen3 VL 32B Instruct pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+24.4points ahead
- Reasons across sets of long documentsAA-LCR+17.6points ahead
- Harder college exam questions with imagesMMMU-Pro+17points ahead
Where Qwen3 VL 8B Instruct pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+5.7points ahead
- Visual perception tasks people solve at a glanceBLINK+1.8points ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+1.3points ahead
Every test, side by side
All 48 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 12.730.117.4+12.7
- Terminal-Bench HardQwen3 VL 32B Instruct by 68.32.3+6
- LiveCodeBench v6Qwen3 VL 32B Instruct by 4.543.839.3+4.5
AgentsQwen3 VL 8B Instruct
- GDPValQwen3 VL 8B Instruct by 9.109.1+9.1
- OSWorld-VerifiedQwen3 VL 8B Instruct by 1.332.633.9+1.3
ReasoningQwen3 VL 32B Instruct
- GPQA DiamondQwen3 VL 32B Instruct by 24.467.142.7+24.4
- Humanity's Last ExamQwen3 VL 32B Instruct by 4.16.82.7+4.1
- CritPttie00tie
FactsQwen3 VL 8B Instruct
- AA-Omniscience · AccuracyQwen3 VL 8B Instruct by 5.714.820.5+5.7
- AA-Omniscience · Non-hallucinationtie8.29tie
Images and chartsQwen3 VL 32B Instruct
Long documentsQwen3 VL 32B Instruct
- AA-LCRQwen3 VL 32B Instruct by 17.634.316.7+17.6
Following instructionsQwen3 VL 32B Instruct
- IFBenchQwen3 VL 32B Instruct by 6.939.232.3+6.9
Other results30 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 Instruct by 20.366.245.9+20.3
- AA-OmniscienceQwen3 VL 8B Instruct by 11.5-63.4-51.9+11.5
- SuperGPQAQwen3 VL 32B Instruct by 10.154.644.5+10.1
- MATH-VisionQwen3 VL 32B Instruct by 9.563.453.9+9.5
- AA Agentic IndexQwen3 VL 8B Instruct by 8.79.718.4+8.7
- Artificial Analysis Coding IndexQwen3 VL 32B Instruct by 8.315.67.3+8.3
- MMLU-ProXQwen3 VL 32B Instruct by 873.465.4+8
- RealWorldQAQwen3 VL 32B Instruct by 7.57971.5+7.5
- IncludeQwen3 VL 32B Instruct by 77467+7
- MMLU-ProQwen3 VL 32B Instruct by 778.671.6+7
- MMMU (val) (Pass@1)Qwen3 VL 32B Instruct by 6.47669.6+6.4
- LVBenchQwen3 VL 32B Instruct by 5.863.858+5.8
- CharadesSTAQwen3 VL 32B Instruct by 5.261.256+5.2
- MMLU-ReduxQwen3 VL 32B Instruct by 4.989.884.9+4.9
- MV-BenchQwen3 VL 32B Instruct by 4.172.868.7+4.1
- BFCL v3Qwen3 VL 32B Instruct by 3.970.266.3+3.9
- InfoVQA_testQwen3 VL 32B Instruct by 3.98783.1+3.9
- AI2DQwen3 VL 32B Instruct by 3.889.585.7+3.8
- ScreenSpot-Pro (No tools)Qwen3 VL 32B Instruct by 3.357.954.6+3.3
- ERQAQwen3 VL 32B Instruct by 348.845.8+3
- HallusionBenchQwen3 VL 32B Instruct by 2.763.861.1+2.7
- OCRBench v2 (Chinese)Qwen3 VL 8B Instruct by 259.261.2+2
- OCRBench v2_enQwen3 VL 32B Instruct by 267.465.4+2
- ScreenSpotQwen3 VL 32B Instruct by 1.495.894.4+1.4
- AA IntelligenceQwen3 VL 32B Instruct by 1.18.47.3+1.1
- IFEvalQwen3 VL 32B Instruct by 184.783.7+1
- DocVQA_testtie96.996.1tie
- MM MTBenchtie8.47.7tie
- CC-OCRtie80.379.9tie
- τ²-Bench Telecom (AA run)tie29.229.2tie
Questions people ask
Which is better, Qwen3 VL 32B Instruct or Qwen3 VL 8B Instruct?
Qwen3 VL 32B Instruct wins five of the seven areas where both have results: coding, reasoning, images and charts, long documents and following instructions. Qwen3 VL 8B Instruct wins agents and facts.
Which is better for coding?
Qwen3 VL 32B Instruct. It wins 3 of the 3 coding tests both models report; Qwen3 VL 8B Instruct wins none.
Which is cheaper?
Qwen3 VL 32B Instruct costs $0.16 per million input tokens and $0.64 per million output tokens; Qwen3 VL 8B Instruct costs $0.18 and $0.70. That makes Qwen3 VL 32B Instruct about 9% cheaper for the same work.
How do you compare the two?
We use the 48 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 30 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.