Qwen3 VL 235B A22B 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 235B A22B Instruct is the stronger all-rounder.
Scores updated · 50 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 235B A22B Instruct6 of 6 tests
- CodingWriting and fixing software30Qwen3 VL 235B A22B Instruct3 of 3 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 235B A22B Instruct2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up10Qwen3 VL 235B A22B Instruct1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts10Qwen3 VL 235B A22B Instruct1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 VL 235B A22B 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 235B A22B Instruct pulls ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+35.3points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+21.8points ahead
- Code for real scientific research problemsSciCode+18.8points ahead
Where Qwen3 VL 4B (Reasoning) pulls ahead
- Real work tasks from 44 professionsGDPVal+7.1points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3 VL 235B A22B Instruct
- SciCodeQwen3 VL 235B A22B Instruct by 18.835.917.1+18.8
- Terminal-Bench HardQwen3 VL 235B A22B Instruct by 5.36.81.5+5.3
- LiveCodeBench v6Qwen3 VL 235B A22B Instruct by 354.351.3+3
AgentsEven
- OSWorld-VerifiedQwen3 VL 235B A22B Instruct by 35.366.731.4+35.3
- GDPValQwen3 VL 4B (Reasoning) by 7.16.713.8+7.1
ReasoningQwen3 VL 235B A22B Instruct
- GPQA DiamondQwen3 VL 235B A22B Instruct by 21.871.249.4+21.8
- Humanity's Last ExamQwen3 VL 235B A22B Instruct by 26.64.6+2
- CritPttie00tie
FactsQwen3 VL 235B A22B Instruct
- AA-Omniscience · AccuracyQwen3 VL 235B A22B Instruct by 8.120.312.2+8.1
- AA-Omniscience · Non-hallucinationtie8.48.2tie
Images and chartsQwen3 VL 235B A22B Instruct
- MMMU-ProQwen3 VL 235B A22B Instruct by 15.667.652+15.6
- CharXiv (RQ)Qwen3 VL 235B A22B Instruct by 11.862.150.3+11.8
- OCRBenchQwen3 VL 235B A22B Instruct by 11.29280.8+11.2
- BLINKQwen3 VL 235B A22B Instruct by 7.370.763.4+7.3
- MathVistaQwen3 VL 235B A22B Instruct by 5.484.979.5+5.4
- MMStarQwen3 VL 235B A22B Instruct by 5.278.473.2+5.2
Long documentsQwen3 VL 235B A22B Instruct
- AA-LCRQwen3 VL 235B A22B Instruct by 11.432.721.3+11.4
Following instructionsQwen3 VL 235B A22B Instruct
- IFBenchQwen3 VL 235B A22B Instruct by 6.142.736.6+6.1
Other results32 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 235B A22B Instruct by 40.677.436.8+40.6
- τ²-Bench Telecom (AA run)Qwen3 VL 235B A22B Instruct by 19.635.115.5+19.6
- AA-OmniscienceQwen3 VL 235B A22B Instruct by 15.7-52.7-68.4+15.7
- IncludeQwen3 VL 235B A22B Instruct by 15.48064.6+15.4
- LVBenchQwen3 VL 235B A22B Instruct by 14.267.753.5+14.2
- SuperGPQAQwen3 VL 235B A22B Instruct by 13.660.446.8+13.6
- MMLU-ProXQwen3 VL 235B A22B Instruct by 12.877.865+12.8
- ScreenSpot-Pro (No tools)Qwen3 VL 235B A22B Instruct by 12.86249.2+12.8
- Artificial Analysis Coding IndexQwen3 VL 235B A22B Instruct by 9.816.56.7+9.8
- CC-OCRQwen3 VL 235B A22B Instruct by 8.482.273.8+8.4
- MMLU-ProQwen3 VL 235B A22B Instruct by 8.281.873.6+8.2
- MMMU (val) (Pass@1)Qwen3 VL 235B A22B Instruct by 7.978.770.8+7.9
- MATH-VisionQwen3 VL 235B A22B Instruct by 6.566.560+6.5
- InfoVQA_testQwen3 VL 235B A22B Instruct by 6.289.283+6.2
- MMLU-ReduxQwen3 VL 235B A22B Instruct by 6.292.286+6.2
- RealWorldQAQwen3 VL 235B A22B Instruct by 6.179.373.2+6.1
- OCRBench v2 (Chinese)Qwen3 VL 235B A22B Instruct by 661.855.8+6
- CharadesSTAQwen3 VL 235B A22B Instruct by 5.864.859+5.8
- OCRBench v2_enQwen3 VL 235B A22B Instruct by 5.367.161.8+5.3
- VideoMMMUQwen3 VL 235B A22B Instruct by 5.374.769.4+5.3
- IFEvalQwen3 VL 235B A22B Instruct by 5.287.882.6+5.2
- AI2DQwen3 VL 235B A22B Instruct by 4.889.784.9+4.8
- AA Agentic IndexQwen3 VL 235B A22B Instruct by 4.719.114.4+4.7
- HMMT 2025Qwen3 VL 235B A22B Instruct by 4.357.453.1+4.3
- ERQAQwen3 VL 235B A22B Instruct by 451.347.3+4
- AA IntelligenceQwen3 VL 235B A22B Instruct by 2.99.97+2.9
- DocVQA_testQwen3 VL 235B A22B Instruct by 2.997.194.2+2.9
- ScreenSpotQwen3 VL 235B A22B Instruct by 2.595.492.9+2.5
- HallusionBenchtie63.264.1tie
- MM MTBenchtie8.57.7tie
- BFCL v3tie67.767.3tie
- AIME 2025tie74.774.5tie
Questions people ask
Which is better, Qwen3 VL 235B A22B Instruct or Qwen3 VL 4B (Reasoning)?
Qwen3 VL 235B A22B 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 235B A22B Instruct. It wins 3 of the 3 coding tests both models report; Qwen3 VL 4B (Reasoning) wins none.
How do you compare the two?
We use the 50 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 32 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.