Qwen3 VL 235B A22B Reasoning vs Qwen3 VL 30B A3B Instruct
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
Wins 0 of 7 areas
—
Qwen3 VL 235B A22B Reasoning is the stronger all-rounder.Qwen3 VL 30B A3B Instruct is cheaper.
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 video51Qwen3 VL 235B A22B Reasoning5 of 7 tests · 1 tie
- CodingWriting and fixing software30Qwen3 VL 235B A22B Reasoning3 of 3 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 235B A22B Reasoning2 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up20Qwen3 VL 235B A22B Reasoning2 of 2 tests
- AgentsCarrying out multi-step tasks on its own10Qwen3 VL 235B A22B Reasoning1 of 1 test
- Long documentsFinding answers in very long texts10Qwen3 VL 235B A22B Reasoning1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 VL 235B A22B Reasoning1 of 1 test
Agents, 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 30B A3B Instruct costs 77% 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 VL 235B A22B Reasoning pulls ahead
- Reasons across sets of long documentsAA-LCR+36.4points ahead
- Recent programming contest problemsLiveCodeBench v6+27.5points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+23.4points ahead
Where Qwen3 VL 30B A3B Instruct pulls ahead
- Reads text in imagesOCRBench+2.8points 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 235B A22B Reasoning
- LiveCodeBench v6Qwen3 VL 235B A22B Reasoning by 27.570.142.6+27.5
- SciCodeQwen3 VL 235B A22B Reasoning by 9.139.930.8+9.1
- Terminal-Bench HardQwen3 VL 235B A22B Reasoning by 5.311.46.1+5.3
AgentsQwen3 VL 235B A22B Reasoning
- OSWorld-VerifiedQwen3 VL 235B A22B Reasoning by 7.838.130.3+7.8
ReasoningQwen3 VL 235B A22B Reasoning
- GPQA DiamondQwen3 VL 235B A22B Reasoning by 7.777.269.5+7.7
- Humanity's Last ExamQwen3 VL 235B A22B Reasoning by 5.611.96.3+5.6
- CritPttie00tie
FactsQwen3 VL 235B A22B Reasoning
- AA-Omniscience · Non-hallucinationQwen3 VL 235B A22B Reasoning by 8.514.86.3+8.5
- AA-Omniscience · AccuracyQwen3 VL 235B A22B Reasoning by 5.320.915.6+5.3
Images and chartsQwen3 VL 235B A22B Reasoning
- CharXiv (RQ)Qwen3 VL 235B A22B Reasoning by 17.266.148.9+17.2
- MMMU-ProQwen3 VL 235B A22B Reasoning by 6.668.762.1+6.6
- MMStarQwen3 VL 235B A22B Reasoning by 6.678.772.1+6.6
- MathVistaQwen3 VL 235B A22B Reasoning by 5.785.880.1+5.7
- Video-MMEQwen3 VL 235B A22B Reasoning by 4.57974.5+4.5
- OCRBenchQwen3 VL 30B A3B Instruct by 2.887.590.3+2.8
- BLINKtie67.167.7tie
Long documentsQwen3 VL 235B A22B Reasoning
- AA-LCRQwen3 VL 235B A22B Reasoning by 36.463.727.3+36.4
Following instructionsQwen3 VL 235B A22B Reasoning
- IFBenchQwen3 VL 235B A22B Reasoning by 23.456.533.1+23.4
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.
- τ²-Bench Telecom (AA run)Qwen3 VL 235B A22B Reasoning by 35.154.119+35.1
- HMMT 2025Qwen3 VL 235B A22B Reasoning by 26.877.450.6+26.8
- AIME 2025Qwen3 VL 235B A22B Reasoning by 20.489.769.3+20.4
- SimpleQAQwen3 VL 235B A22B Reasoning by 17.444.427+17.4
- AA-OmniscienceQwen3 VL 235B A22B Reasoning by 16.9-46.5-63.4+16.9
- MATH-VisionQwen3 VL 235B A22B Reasoning by 14.474.660.2+14.4
- VideoMMMUQwen3 VL 235B A22B Reasoning by 11.38068.7+11.3
- SuperGPQAQwen3 VL 235B A22B Reasoning by 11.264.353.1+11.2
- MMLU-ProXQwen3 VL 235B A22B Reasoning by 9.780.670.9+9.7
- ERQAQwen3 VL 235B A22B Reasoning by 9.552.543+9.5
- IncludeQwen3 VL 235B A22B Reasoning by 8.48071.6+8.4
- RealWorldQAQwen3 VL 235B A22B Reasoning by 7.681.373.7+7.6
- InfoVQA_testQwen3 VL 235B A22B Reasoning by 7.589.582+7.5
- MMMU (val) (Pass@1)Qwen3 VL 235B A22B Reasoning by 6.480.674.2+6.4
- MMLU-ProQwen3 VL 235B A22B Reasoning by 683.877.8+6
- OCRBench v2 (Chinese)Qwen3 VL 235B A22B Reasoning by 5.763.557.8+5.7
- BFCL v3Qwen3 VL 235B A22B Reasoning by 5.671.966.3+5.6
- AA IntelligenceQwen3 VL 235B A22B Reasoning by 5.513.47.9+5.5
- MMLU-ReduxQwen3 VL 235B A22B Reasoning by 5.393.788.4+5.3
- HallusionBenchQwen3 VL 235B A22B Reasoning by 5.266.761.5+5.2
- AI2DQwen3 VL 235B A22B Reasoning by 4.289.285+4.2
- OCRBench v2_enQwen3 VL 235B A22B Reasoning by 3.666.863.2+3.6
- IFEvalQwen3 VL 235B A22B Reasoning by 2.488.285.8+2.4
- DocVQA_testQwen3 VL 235B A22B Reasoning by 1.596.595+1.5
- ScreenSpot-Pro (No tools)Qwen3 VL 235B A22B Reasoning by 1.361.860.5+1.3
- LVBenchQwen3 VL 235B A22B Reasoning by 1.163.662.5+1.1
- CC-OCRtie81.580.7tie
- ScreenSpottie95.494.7tie
- MM MTBenchtie8.58.1tie
- CharadesSTAtie63.563.5tie
Questions people ask
Which is better, Qwen3 VL 235B A22B Reasoning or Qwen3 VL 30B A3B Instruct?
Qwen3 VL 235B A22B Reasoning wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, long documents and following instructions. Qwen3 VL 30B A3B Instruct wins none, but costs 77% less.
Which is better for coding?
Qwen3 VL 235B A22B Reasoning. It wins 3 of the 3 coding tests both models report; Qwen3 VL 30B A3B Instruct wins none.
Which is cheaper?
Qwen3 VL 235B A22B Reasoning costs $0.40 per million input tokens and $4.00 per million output tokens; Qwen3 VL 30B A3B Instruct costs $0.20 and $0.80. That makes Qwen3 VL 30B A3B Instruct about 77% 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.