Qwen3 VL 235B A22B Instruct vs Qwen3 VL 32B Reasoning
Wins 1 of 7 areas
Agents
Wins 3 of 7 areas
Reasoning · Long documents · Following instructions
Qwen3 VL 32B Reasoning wins more areas, narrowly.Qwen3 VL 235B A22B Instruct is better at agents.
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.
- ReasoningHard problems that need careful thinking02Qwen3 VL 32B Reasoning2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts01Qwen3 VL 32B Reasoning1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3 VL 32B Reasoning1 of 1 test
- AgentsCarrying out multi-step tasks on its own10Qwen3 VL 235B A22B Instruct1 of 2 tests · 1 tie
- CodingWriting and fixing software11Even1 each · 1 tie
- FactsGetting facts right instead of making them up11Even1 each
- Images and chartsUnderstanding pictures, charts and video33Even3 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.
Qwen3 VL 32B Reasoning costs 60% 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 Reasoning pulls ahead
- Reasons across sets of long documentsAA-LCR+22.6points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+16.7points ahead
- Recent programming contest problemsLiveCodeBench v6+11.3points ahead
Where Qwen3 VL 235B A22B Instruct pulls ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+25.7points ahead
- Code for real scientific research problemsSciCode+7.4points ahead
- Reads text in imagesOCRBench+6.5points 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.
CodingEven
- LiveCodeBench v6Qwen3 VL 32B Reasoning by 11.354.365.6+11.3
- SciCodeQwen3 VL 235B A22B Instruct by 7.435.928.5+7.4
- Terminal-Bench Hardtie6.87.6tie
AgentsQwen3 VL 235B A22B Instruct
- OSWorld-VerifiedQwen3 VL 235B A22B Instruct by 25.766.741+25.7
- GDPValtie6.77.3tie
ReasoningQwen3 VL 32B Reasoning
- Humanity's Last ExamQwen3 VL 32B Reasoning by 3.56.610.1+3.5
- GPQA DiamondQwen3 VL 32B Reasoning by 2.171.273.3+2.1
- CritPttie00tie
FactsEven
- AA-Omniscience · Non-hallucinationQwen3 VL 32B Reasoning by 7.98.416.3+7.9
- AA-Omniscience · AccuracyQwen3 VL 235B A22B Instruct by 3.320.317+3.3
Images and chartsEven
Long documentsQwen3 VL 32B Reasoning
- AA-LCRQwen3 VL 32B Reasoning by 22.632.755.3+22.6
Following instructionsQwen3 VL 32B Reasoning
- IFBenchQwen3 VL 32B Reasoning by 16.742.759.4+16.7
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 235B A22B Instruct by 16.977.460.5+16.9
- τ²-Bench Telecom (AA run)Qwen3 VL 32B Reasoning by 10.535.145.6+10.5
- AIME 2025Qwen3 VL 32B Reasoning by 974.783.7+9
- LVBenchQwen3 VL 235B A22B Instruct by 5.167.762.6+5.1
- ScreenSpot-Pro (No tools)Qwen3 VL 235B A22B Instruct by 4.96257.1+4.9
- AA Agentic IndexQwen3 VL 32B Reasoning by 4.319.123.4+4.3
- VideoMMMUQwen3 VL 32B Reasoning by 4.374.779+4.3
- HallusionBenchQwen3 VL 32B Reasoning by 4.263.267.4+4.2
- BFCL v3Qwen3 VL 32B Reasoning by 467.771.7+4
- IncludeQwen3 VL 235B A22B Instruct by 3.78076.3+3.7
- MATH-VisionQwen3 VL 32B Reasoning by 3.766.570.2+3.7
- SimpleQAQwen3 VL 32B Reasoning by 3.551.955.4+3.5
- AA IntelligenceQwen3 VL 32B Reasoning by 29.911.9+2
- Artificial Analysis Coding IndexQwen3 VL 235B A22B Instruct by 216.514.5+2
- CharadesSTAQwen3 VL 235B A22B Instruct by 264.862.8+2
- SuperGPQAQwen3 VL 235B A22B Instruct by 1.460.459+1.4
- OCRBench v2_enQwen3 VL 32B Reasoning by 1.367.168.4+1.3
- DocVQA_testQwen3 VL 235B A22B Instruct by 197.196.1+1
- ERQAQwen3 VL 32B Reasoning by 151.352.3+1
- RealWorldQAtie79.378.4tie
- AI2Dtie89.788.9tie
- MMLU-ProXtie77.877.2tie
- MMMU (val) (Pass@1)tie78.778.1tie
- MMLU-Protie81.882.1tie
- MMLU-Reduxtie92.291.9tie
- OCRBench v2 (Chinese)tie61.862.1tie
- ScreenSpottie95.495.7tie
- AA-Omnisciencetie-52.7-52.5tie
- MM MTBenchtie8.58.3tie
- IFEvaltie87.887.8tie
- InfoVQA_testtie89.289.2tie
Questions people ask
Which is better, Qwen3 VL 235B A22B Instruct or Qwen3 VL 32B Reasoning?
Qwen3 VL 32B Reasoning wins three of the seven areas where both have results: reasoning, long documents and following instructions. Qwen3 VL 235B A22B Instruct wins agents. They are level on coding, facts and images and charts.
Which is better for coding?
Neither. They win 1 coding test each of the 3 both models report, and 1 is a tie.
Which is cheaper?
Qwen3 VL 235B A22B Instruct costs $0.40 per million input tokens and $1.60 per million output tokens; Qwen3 VL 32B Reasoning costs $0.16 and $0.64. That makes Qwen3 VL 32B Reasoning about 60% cheaper for the same work.
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.