Qwen3 VL 8B Thinking
Qwen3-VL is a large multimodal model that unifies vision, language, and reasoning to achieve human-level perception and cognition across text, images, and video. Built on a 235B-parameter architecture, it integrates early joint training of visual and textual modalities for strong language grounding. The model supports up to a 1 million-token context window and excels at visual understanding, spatial reasoning, long video comprehension, and tool-based interaction. It can generate code from images, perform precise 2D/3D object grounding, and operate digital interfaces like a visual agent. The “Instruct” version rivals Gemini 2.5 Pro in perception benchmarks, while the “Thinking” version leads in multimodal reasoning and STEM tasks. With multilingual OCR, creative writing, and fine-grained scene interpretation, Qwen3-VL establishes a new open-source frontier for integrated vision-language intelligence.
Benchmark results
| Benchmark | Score | Tags | Source |
|---|---|---|---|
| AI2D | 84.9% | self-reported llm-stats | link → |
| AIME 2025 | 80.3% | self-reported llm-stats | link → |
| Arena-Hard v2 | 51.1% | self-reported llm-stats | link → |
| BFCL-v3 | 63.0% | self-reported llm-stats | link → |
| BLINK | 68.7% | self-reported llm-stats | link → |
| CC-OCR | 76.3% | self-reported llm-stats | link → |
| CharadesSTA | 59.9% | self-reported llm-stats | link → |
| CharXiv-D | 85.9% | self-reported llm-stats | link → |
| CharXiv-R | 53.0% | self-reported llm-stats | link → |
| Creative Writing v3 | 0.824 | self-reported llm-stats | link → |
| DocVQAtest | 95.3% | self-reported llm-stats | link → |
| ERQA | 46.8% | self-reported llm-stats | link → |
| GPQA | 69.9% | self-reported llm-stats | link → |
| Hallusion Bench | 65.4% | self-reported llm-stats | link → |
| HMMT25 | 60.6% | self-reported llm-stats | link → |
| IFEval | 83.2% | self-reported llm-stats | link → |
| Include | 69.5% | self-reported llm-stats | link → |
| InfoVQAtest | 86.0% | self-reported llm-stats | link → |
| LiveBench 20241125 | 69.8% | self-reported llm-stats | link → |
| LiveCodeBench v6 | 58.6% | self-reported llm-stats | link → |
| LVBench | 55.8% | self-reported llm-stats | link → |
| MathVision | 62.7% | self-reported llm-stats | link → |
| MathVista-Mini | 81.4% | self-reported llm-stats | link → |
| MLVU-M | 75.1% | self-reported llm-stats | link → |
| MM-MT-Bench | 8 | self-reported llm-stats | link → |
| MMBench-V1.1 | 87.5% | self-reported llm-stats | link → |
| MMLU | 85.2% | self-reported llm-stats | link → |
| MMLU-Pro | 77.3% | self-reported llm-stats | link → |
| MMLU-ProX | 70.7% | self-reported llm-stats | link → |
| MMLU-Redux | 88.8% | self-reported llm-stats | link → |
| MMMU (val) | 74.1% | self-reported llm-stats | link → |
| MMMU-Pro | 60.4% | self-reported llm-stats | link → |
| MMStar | 75.3% | self-reported llm-stats | link → |
| MuirBench | 76.8% | self-reported llm-stats | link → |
| Multi-IF | 75.1% | self-reported llm-stats | link → |
| MVBench | 69.0% | self-reported llm-stats | link → |
| OCRBench | 81.9% | self-reported llm-stats | link → |
| OCRBench-V2 (en) | 63.9% | self-reported llm-stats | link → |
| OCRBench-V2 (zh) | 59.2% | self-reported llm-stats | link → |
| ODinW | 39.8% | self-reported llm-stats | link → |
| OSWorld | 33.9% | self-reported llm-stats | link → |
| PolyMATH | 47.5% | self-reported llm-stats | link → |
| RealWorldQA | 73.5% | self-reported llm-stats | link → |
| ScreenSpot | 93.6% | self-reported llm-stats | link → |
| ScreenSpot Pro | 46.6% | self-reported llm-stats | link → |
| SimpleQA | 49.6% | self-reported llm-stats | link → |
| SuperGPQA | 51.2% | self-reported llm-stats | link → |
| Video-MME | 71.8% | self-reported llm-stats | link → |
| VideoMMMU | 72.8% | self-reported llm-stats | link → |
| WritingBench | 85.5% | self-reported llm-stats | link → |