Loading…
Loading…
Models / Qwen2.5-VL
4 models (3B to 72B) · ↓ 12.6M downloads a month · from $0.10/h on RTX 3060 right now. Each model's versions do the same job; quantized ones (AWQ, FP8, 4-bit) need less GPU memory and run on cheaper machines.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalQwen/Qwen2.5-VL-7B-Instruct | vLLM | 17 GB | 21.2 GB | $0.18/h 2x RTX 3060 |
| AWQQwen/Qwen2.5-VL-7B-Instruct-AWQ | vLLM | 6.9 GB | 10.4 GB | $0.10/h RTX 3060 |
| NVFP4 (nvidia)nvidia/Qwen2.5-VL-7B-Instruct-NVFP4 | vLLM | 7.2 GB | 10.7 GB | $0.10/h RTX 3060 |
| olmOCR 7B 0825 FP8 (allenai)allenai/olmOCR-7B-0825-FP8 | vLLM | 10 GB | 13.9 GB | $0.14/h RTX A4000 |
| olmOCR 2 7B 1025 FP8 (allenai)allenai/olmOCR-2-7B-1025-FP8 | vLLM | 10 GB | 13.9 GB | $0.14/h RTX A4000 |
| FP8 Dynamic (RedHatAI)RedHatAI/Qwen2.5-VL-7B-Instruct-FP8-Dynamic | vLLM | 10 GB | 13.9 GB | $0.14/h RTX A4000 |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| OriginalQwen/Qwen2.5-VL-72B-Instruct | vLLM | 147 GB | 168.3 GB | $1.38/h 8x RTX 3090 |
Model details come from Hugging Face and are refreshed daily; prices follow the live offers. GPU memory is GridPup's estimate: the weights, room for an 8,192-token conversation (chat models) and the runtime's own memory. Large models are split across the GPUs of one machine ("4x RTX 4090").