Qwen 3 VL 4B (Thinking): benchmark results
Alibaba Qwen 3 VL 4B vision-language model evaluated with thinking enabled. Provider: Alibaba. Released 2025-07-01. Access: Open.
Unified ELO 1446 ± 1, rank #1123 of 1761 rated models, from 85 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| OCR-Robust | 78.63 | OCR1.0 Clean (self-reported) | 82.4 |
| MedLayXPlain | 63.3 | S (self-reported) | 74.2 |
| LLM Stats (MuirBench) | 75 | Score (%) | 63.6 |
| K-MetBench | 66.1 | Accuracy (self-reported) | 62.1 |
| Video-MMMU | 69.4 | Overall Score | 55.1 |
| LLM Stats (Hallusion Bench) | 64.1 | Score (%) | 52.9 |
| AA Omniscience - Software Engineering (SWE) - Swift | 32 | Accuracy (%) | 51.2 |
| LLM Stats (MMStar) | 73.2 | Score (%) | 50 |
| LLM Stats (Multi-IF) | 73.6 | Score (%) | 50 |
| AA Omniscience - Software Engineering (SWE) - JavaScript | 27.27 | Accuracy (%) | 48.3 |
| RealWorldQA | 73.2 | RealWorldQA (self-reported) | 47.2 |
| AA Omniscience - Software Engineering (SWE) - Dart | 16 | Accuracy (%) | 43.7 |
Interactive version: theaggregate.ai/model?slug=qwen-3-vl-4b-thinking · How It Works · Data refreshed daily, snapshot 2026-09-05.