Qwen3 VL 30B A3B Reasoning vs Qwen3 VL Thinking (8B)
Wins 5 of 7 areas
Coding · Reasoning · Images and charts · Long documents · Following instructions
Wins 1 of 7 areas
Agents
Qwen3 VL 30B A3B Reasoning is the stronger all-rounder.Qwen3 VL Thinking (8B) is cheaper and better at agents.
Scores updated · 54 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 30B A3B Reasoning4 of 7 tests · 2 ties
- CodingWriting and fixing software30Qwen3 VL 30B A3B Reasoning3 of 3 tests
- ReasoningHard problems that need careful thinking20Qwen3 VL 30B A3B Reasoning2 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts10Qwen3 VL 30B A3B Reasoning1 of 1 test
- Following instructionsDoing exactly what it is asked10Qwen3 VL 30B A3B Reasoning1 of 1 test
- AgentsCarrying out multi-step tasks on its own01Qwen3 VL Thinking (8B)1 of 2 tests · 1 tie
- FactsGetting facts right instead of making them up11Even1 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 Thinking (8B) costs 12% 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 30B A3B Reasoning pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+14.1points ahead
- Reasons across sets of long documentsAA-LCR+9.4points ahead
- Code for real scientific research problemsSciCode+6.9points ahead
Where Qwen3 VL Thinking (8B) pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+3.5points ahead
- Completes tasks by operating a computer desktopOSWorld-Verified+3.3points ahead
- Visual perception tasks people solve at a glanceBLINK+3.3points ahead
Every test, side by side
All 54 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingQwen3 VL 30B A3B Reasoning
- SciCodeQwen3 VL 30B A3B Reasoning by 6.928.821.9+6.9
- LiveCodeBench v6Qwen3 VL 30B A3B Reasoning by 5.664.258.6+5.6
- Terminal-Bench HardQwen3 VL 30B A3B Reasoning by 1.55.33.8+1.5
AgentsQwen3 VL Thinking (8B)
- OSWorld-VerifiedQwen3 VL Thinking (8B) by 3.330.633.9+3.3
- GDPValtie8.38.4tie
ReasoningQwen3 VL 30B A3B Reasoning
- GPQA DiamondQwen3 VL 30B A3B Reasoning by 14.17257.9+14.1
- Humanity's Last ExamQwen3 VL 30B A3B Reasoning by 5.18.93.8+5.1
- CritPttie00.3tie
FactsEven
- AA-Omniscience · AccuracyQwen3 VL Thinking (8B) by 3.516.920.4+3.5
- AA-Omniscience · Non-hallucinationQwen3 VL 30B A3B Reasoning by 1.610.28.6+1.6
Images and chartsQwen3 VL 30B A3B Reasoning
- MMMU-ProQwen3 VL 30B A3B Reasoning by 5.261.856.6+5.2
- CharXiv (RQ)Qwen3 VL 30B A3B Reasoning by 3.656.653+3.6
- BLINKQwen3 VL Thinking (8B) by 3.365.468.7+3.3
- OCRBenchQwen3 VL 30B A3B Reasoning by 283.981.9+2
- Video-MMEQwen3 VL 30B A3B Reasoning by 1.573.371.8+1.5
- MathVistatie81.981.4tie
- MMStartie75.575.3tie
Long documentsQwen3 VL 30B A3B Reasoning
- AA-LCRQwen3 VL 30B A3B Reasoning by 9.442.733.3+9.4
Following instructionsQwen3 VL 30B A3B Reasoning
- IFBenchQwen3 VL 30B A3B Reasoning by 5.245.139.9+5.2
Other results35 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.
- SimpleQAQwen3 VL Thinking (8B) by 25.723.949.6+25.7
- ScreenSpot-Pro (No tools)Qwen3 VL 30B A3B Reasoning by 10.757.346.6+10.7
- HMMT 2025Qwen3 VL 30B A3B Reasoning by 767.660.6+7
- Arena-Hard-V2Qwen3 VL 30B A3B Reasoning by 5.656.751.1+5.6
- BFCL v3Qwen3 VL 30B A3B Reasoning by 5.668.663+5.6
- MMLU-ProXQwen3 VL 30B A3B Reasoning by 5.476.170.7+5.4
- AA-OmniscienceQwen3 VL Thinking (8B) by 5.3-57.6-52.3+5.3
- SuperGPQAQwen3 VL 30B A3B Reasoning by 5.256.451.2+5.2
- IncludeQwen3 VL 30B A3B Reasoning by 574.569.5+5
- RealWorldQAQwen3 VL 30B A3B Reasoning by 3.977.473.5+3.9
- LVBenchQwen3 VL 30B A3B Reasoning by 3.459.255.8+3.4
- Artificial Analysis Coding IndexQwen3 VL 30B A3B Reasoning by 3.313.19.8+3.3
- MMLU-ProQwen3 VL 30B A3B Reasoning by 3.280.577.3+3.2
- MATH-VisionQwen3 VL 30B A3B Reasoning by 365.762.7+3
- MV-BenchQwen3 VL 30B A3B Reasoning by 37269+3
- AIME 2025Qwen3 VL 30B A3B Reasoning by 2.883.180.3+2.8
- CharadesSTAQwen3 VL 30B A3B Reasoning by 2.862.759.9+2.8
- τ²-Bench Telecom (AA run)Qwen3 VL Thinking (8B) by 2.619.922.5+2.6
- VideoMMMUQwen3 VL 30B A3B Reasoning by 2.27572.8+2.2
- MMLU-ReduxQwen3 VL 30B A3B Reasoning by 2.190.988.8+2.1
- AI2DQwen3 VL 30B A3B Reasoning by 286.984.9+2
- MMMU (val) (Pass@1)Qwen3 VL 30B A3B Reasoning by 1.97674.1+1.9
- CC-OCRQwen3 VL 30B A3B Reasoning by 1.577.876.3+1.5
- ERQAQwen3 VL Thinking (8B) by 1.545.346.8+1.5
- IFEvalQwen3 VL Thinking (8B) by 1.581.783.2+1.5
- AA IntelligenceQwen3 VL 30B A3B Reasoning by 1.39.58.2+1.3
- OCRBench v2_enQwen3 VL Thinking (8B) by 1.362.663.9+1.3
- OCRBench v2 (Chinese)Qwen3 VL 30B A3B Reasoning by 1.260.459.2+1.2
- ScreenSpotQwen3 VL 30B A3B Reasoning by 1.194.793.6+1.1
- AA Agentic Indextie14.915.6tie
- HallusionBenchtie6665.4tie
- DocVQA_testtie9595.3tie
- Creative Writing v3tie82.582.4tie
- MM MTBenchtie7.98tie
- InfoVQA_testtie8686tie
Questions people ask
Which is better, Qwen3 VL 30B A3B Reasoning or Qwen3 VL Thinking (8B)?
Qwen3 VL 30B A3B Reasoning wins five of the seven areas where both have results: coding, reasoning, images and charts, long documents and following instructions. Qwen3 VL Thinking (8B) wins agents, and costs 12% less. They are level on facts.
Which is better for coding?
Qwen3 VL 30B A3B Reasoning. It wins 3 of the 3 coding tests both models report; Qwen3 VL Thinking (8B) wins none.
Which is cheaper?
Qwen3 VL 30B A3B Reasoning costs $0.20 per million input tokens and $2.40 per million output tokens; Qwen3 VL Thinking (8B) costs $0.18 and $2.10. That makes Qwen3 VL Thinking (8B) about 12% cheaper for the same work.
How do you compare the two?
We use the 54 benchmark tests both models have published scores on. The verdict counts the 19 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 35 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.