Qwen3 VL 8B Instruct
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 | 85.7% | self-reported llm-stats | link → |
| AIME 2025 | 45.9% | self-reported llm-stats | link → |
| BFCL-v3 | 66.3% | self-reported llm-stats | link → |
| BLINK | 69.1% | self-reported llm-stats | link → |
| CC-OCR | 79.9% | self-reported llm-stats | link → |
| CharadesSTA | 56.0% | self-reported llm-stats | link → |
| CharXiv-D | 83.0% | self-reported llm-stats | link → |
| CharXiv-R | 46.4% | self-reported llm-stats | link → |
| DocVQAtest | 96.1% | self-reported llm-stats | link → |
| ERQA | 45.8% | self-reported llm-stats | link → |
| Hallusion Bench | 61.1% | self-reported llm-stats | link → |
| HMMT25 | 32.5% | self-reported llm-stats | link → |
| IFEval | 83.7% | self-reported llm-stats | link → |
| Include | 67.0% | self-reported llm-stats | link → |
| InfoVQAtest | 83.1% | self-reported llm-stats | link → |
| LiveBench 20241125 | 62.0% | self-reported llm-stats | link → |
| LiveCodeBench v6 | 39.3% | self-reported llm-stats | link → |
| LVBench | 58.0% | self-reported llm-stats | link → |
| MathVision | 53.9% | self-reported llm-stats | link → |
| MathVista-Mini | 77.2% | self-reported llm-stats | link → |
| MLVU-M | 78.1% | self-reported llm-stats | link → |
| MM-MT-Bench | 7.7 | self-reported llm-stats | link → |
| MMBench-V1.1 | 85.0% | self-reported llm-stats | link → |
| MMLU | 80.7% | self-reported llm-stats | link → |
| MMLU-Pro | 71.6% | self-reported llm-stats | link → |
| MMLU-ProX | 65.4% | self-reported llm-stats | link → |
| MMLU-Redux | 84.9% | self-reported llm-stats | link → |
| MMMU (val) | 69.6% | self-reported llm-stats | link → |
| MMMU-Pro | 55.9% | self-reported llm-stats | link → |
| MMStar | 70.9% | self-reported llm-stats | link → |
| MuirBench | 64.4% | self-reported llm-stats | link → |
| Multi-IF | 75.1% | self-reported llm-stats | link → |
| MVBench | 68.7% | self-reported llm-stats | link → |
| OCRBench | 89.6% | self-reported llm-stats | link → |
| OCRBench-V2 (en) | 65.4% | self-reported llm-stats | link → |
| OCRBench-V2 (zh) | 61.2% | self-reported llm-stats | link → |
| ODinW | 44.7% | self-reported llm-stats | link → |
| OSWorld | 33.9% | self-reported llm-stats | link → |
| PolyMATH | 30.4% | self-reported llm-stats | link → |
| RealWorldQA | 71.5% | self-reported llm-stats | link → |
| ScreenSpot | 94.4% | self-reported llm-stats | link → |
| ScreenSpot Pro | 54.6% | self-reported llm-stats | link → |
| SuperGPQA | 44.5% | self-reported llm-stats | link → |
| Video-MME | 71.4% | self-reported llm-stats | link → |
| VideoMMMU | 65.3% | self-reported llm-stats | link → |
| WritingBench | 83.1% | self-reported llm-stats | link → |