Qwen3 VL 4B Instruct vs Qwen3 VL 4B (Reasoning)
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Long documents · Following instructions
Qwen3 VL 4B (Reasoning) 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.
- CodingWriting and fixing software03Qwen3 VL 4B (Reasoning)3 of 3 tests
- Images and chartsUnderstanding pictures, charts and video24Qwen3 VL 4B (Reasoning)4 of 6 tests
- ReasoningHard problems that need careful thinking02Qwen3 VL 4B (Reasoning)2 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own02Qwen3 VL 4B (Reasoning)2 of 2 tests
- FactsGetting facts right instead of making them up02Qwen3 VL 4B (Reasoning)2 of 2 tests
- Long documentsFinding answers in very long texts01Qwen3 VL 4B (Reasoning)1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3 VL 4B (Reasoning)1 of 1 test
Long documents and following instructions rest on a single test each.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Qwen3 VL 4B (Reasoning) pulls ahead
- Recent programming contest problemsLiveCodeBench v6+13.4points ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+12.3points ahead
- Reasoning about charts from research papersCharXiv (RQ)+10.6points 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 4B (Reasoning)
- LiveCodeBench v6Qwen3 VL 4B (Reasoning) by 13.437.951.3+13.4
- SciCodeQwen3 VL 4B (Reasoning) by 3.413.717.1+3.4
- Terminal-Bench HardQwen3 VL 4B (Reasoning) by 1.501.5+1.5
AgentsQwen3 VL 4B (Reasoning)
- GDPValQwen3 VL 4B (Reasoning) by 13.8013.8+13.8
- OSWorld-VerifiedQwen3 VL 4B (Reasoning) by 5.226.231.4+5.2
ReasoningQwen3 VL 4B (Reasoning)
- GPQA DiamondQwen3 VL 4B (Reasoning) by 12.337.149.4+12.3
- Humanity's Last ExamQwen3 VL 4B (Reasoning) by 13.64.6+1
- CritPttie00tie
FactsQwen3 VL 4B (Reasoning)
- AA-Omniscience · Non-hallucinationQwen3 VL 4B (Reasoning) by 5.62.68.2+5.6
- AA-Omniscience · AccuracyQwen3 VL 4B (Reasoning) by 1.310.912.2+1.3
Images and chartsQwen3 VL 4B (Reasoning)
Long documentsQwen3 VL 4B (Reasoning)
- AA-LCRQwen3 VL 4B (Reasoning) by 7.31421.3+7.3
Following instructionsQwen3 VL 4B (Reasoning)
- IFBenchQwen3 VL 4B (Reasoning) by 4.831.836.6+4.8
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.
- AIME 2025Qwen3 VL 4B (Reasoning) by 27.946.674.5+27.9
- HMMT 2025Qwen3 VL 4B (Reasoning) by 22.430.753.1+22.4
- VideoMMMUQwen3 VL 4B (Reasoning) by 13.256.269.4+13.2
- ScreenSpot-Pro (No tools)Qwen3 VL 4B Instruct by 10.359.549.2+10.3
- MATH-VisionQwen3 VL 4B (Reasoning) by 8.451.660+8.4
- τ²-Bench Telecom (AA run)Qwen3 VL 4B Instruct by 7.923.415.5+7.9
- AA-OmniscienceQwen3 VL 4B (Reasoning) by 7.6-76-68.4+7.6
- AA Agentic IndexQwen3 VL 4B (Reasoning) by 6.67.814.4+6.6
- HallusionBenchQwen3 VL 4B (Reasoning) by 6.557.664.1+6.5
- MMLU-ProQwen3 VL 4B (Reasoning) by 6.567.173.6+6.5
- SuperGPQAQwen3 VL 4B (Reasoning) by 6.540.346.8+6.5
- ERQAQwen3 VL 4B (Reasoning) by 641.347.3+6
- MMLU-ProXQwen3 VL 4B (Reasoning) by 5.659.465+5.6
- MMLU-ReduxQwen3 VL 4B (Reasoning) by 4.581.586+4.5
- BFCL v3Qwen3 VL 4B (Reasoning) by 463.367.3+4
- CharadesSTAQwen3 VL 4B (Reasoning) by 3.555.559+3.5
- MMMU (val) (Pass@1)Qwen3 VL 4B (Reasoning) by 3.467.470.8+3.4
- IncludeQwen3 VL 4B (Reasoning) by 3.261.464.6+3.2
- InfoVQA_testQwen3 VL 4B (Reasoning) by 2.780.383+2.7
- LVBenchQwen3 VL 4B Instruct by 2.756.253.5+2.7
- CC-OCRQwen3 VL 4B Instruct by 2.476.273.8+2.4
- RealWorldQAQwen3 VL 4B (Reasoning) by 2.370.973.2+2.3
- Artificial Analysis Coding IndexQwen3 VL 4B (Reasoning) by 2.14.66.7+2.1
- OCRBench v2_enQwen3 VL 4B Instruct by 1.963.761.8+1.9
- OCRBench v2 (Chinese)Qwen3 VL 4B Instruct by 1.857.655.8+1.8
- AA IntelligenceQwen3 VL 4B (Reasoning) by 1.35.77+1.3
- DocVQA_testQwen3 VL 4B Instruct by 1.195.394.2+1.1
- ScreenSpotQwen3 VL 4B Instruct by 1.19492.9+1.1
- AI2Dtie84.184.9tie
- MV-Benchtie68.969.3tie
- IFEvaltie82.382.6tie
- MM MTBenchtie7.57.7tie
Questions people ask
Which is better, Qwen3 VL 4B Instruct or Qwen3 VL 4B (Reasoning)?
Qwen3 VL 4B (Reasoning) wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, long documents and following instructions. Qwen3 VL 4B Instruct wins none.
Which is better for coding?
Qwen3 VL 4B (Reasoning). It wins 3 of the 3 coding tests both models report; Qwen3 VL 4B Instruct 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.