family qwen · 16 variants · unknown
| benchmark | category | best score |
|---|---|---|
| DocVQA | multimodal | 96.0 |
| MMBench | multimodal | 87.9 |
| MathVista | multimodal | 73.8 |
| MMMU | multimodal | 52.3 |
| variant | kind | quant | bits | vRAM | dl | org | year |
|---|---|---|---|---|---|---|---|
| Qwen2.5-VL-3B-Instruct | original | bf16 | 16.0 | 7,534,410 | Qwen | 2025 | |
| Qwen2.5-VL-3B-Instruct-AWQ | awq | awq | 2.5 | 351,921 | Qwen | 2025 | |
| Qwen2.5-VL-3B-Instruct-GGUF | gguf | gguf | 4.0 | 18,523 | ggml-org | 2025 | |
| Qwen2.5-VL-3B-Instruct-GGUF | gguf | gguf | 3.0 | 15,317 | lmstudio-community | 2025 | |
| Qwen2.5-VL-3B-Instruct-unsloth-bnb-4bit | bnb-4bit | 4bit | 7,851 | unsloth | 2025 | ||
| Qwen2.5-VL-3B-Instruct-GGUF | gguf | gguf | 16.0 | 7,375 | unsloth | 2025 | |
| Qwen2.5-VL-3B-Instruct-GGUF | gguf | gguf | 16.0 | 6,539 | Mungert | 2025 | |
| Qwen2.5-VL-3B-Instruct-bnb-4bit | bnb-4bit | 4bit | 3,834 | unsloth | 2025 | ||
| Qwen2.5-VL-3B-Instruct | original | bf16 | 16.0 | 2,192 | unsloth | 2025 | |
| NV-Reason-CXR-3B | original | bf16 | 16.0 | 755 | nvidia | 2025 | |
| GUI-Actor-3B-Qwen2.5-VL | original | bf16 | 16.0 | 413 | microsoft | 2025 | |
| OCRFlux-3B-GGUF | gguf | gguf | 16.0 | 145 | Mungert | 2025 | |
| Nanonets-OCR2-3B-GGUF | gguf | gguf | 16.0 | 128 | Mungert | 2025 | |
| Holo1-3B-GGUF | gguf | gguf | 16.0 | 108 | Mungert | 2025 | |
| Spatial-SSRL-3B | original | bf16 | 16.0 | 101 | internlm | 2026 | |
| Nanonets-OCR-s-GGUF | gguf | gguf | 16.0 | 96 | Mungert | 2025 |