Models / Qwen1.5
Qwen1.5 by Qwen
15 models (0.5B to 110B) · ↓ 477.2K downloads a month · from $0.08/h on GTX 1660 SUPER 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.
Qwen1.5 0.5B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-0.5B-Chat | vLLM | 1.2 GB | 4.5 GB | $0.10/h RTX 3060 |
| Main (Xenova)Xenova/Qwen1.5-0.5B-Chat | ONNX Runtime | 480 MB | 1.7 GB | $0.08/h GTX 1660 SUPER |
Qwen1.5 0.5B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-0.5B | vLLM | 1.2 GB | 4.5 GB | $0.10/h RTX 3060 |
Qwen1.5 1.8B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-1.8B-Chat | vLLM | 3.7 GB | 8.1 GB | $0.10/h RTX 3060 |
Qwen1.5 1.8B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-1.8B | vLLM | 3.7 GB | 8.1 GB | $0.10/h RTX 3060 |
Qwen1.5 4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-4B-Chat | vLLM | 7.9 GB | 14.7 GB | $0.14/h RTX A4000 |
Qwen1.5 4B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-4B | vLLM | 7.9 GB | 14.7 GB | $0.14/h RTX A4000 |
Qwen1.5 7B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-7B | vLLM | 15 GB | 24.2 GB | $0.34/h 2x RTX 5060 Ti |
Qwen1.5 7B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-7B-Chat | vLLM | 15 GB | 24.2 GB | $0.34/h 2x RTX 5060 Ti |
Qwen1.5 14B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-14B-Chat | vLLM | 28 GB | 41.2 GB | $0.45/h Nvidia A40 |
Qwen1.5 14B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-14B | vLLM | 28 GB | 41.2 GB | $0.45/h Nvidia A40 |
Qwen1.5 32B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-32B-Chat | vLLM | 65 GB | 76.9 GB | $0.89/h 2x Nvidia A40 |
Qwen1.5 32B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-32B | vLLM | 65 GB | 76.9 GB | $0.89/h 2x Nvidia A40 |
Qwen1.5 72B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-72B | vLLM | 145 GB | 186.7 GB | $2.94/h 2x RTX PRO 6000 |
| ChatQwen/Qwen1.5-72B-Chat | vLLM | 145 GB | 186.7 GB | $2.94/h 2x RTX PRO 6000 |
Qwen1.5 110B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-110B-Chat | vLLM | 222 GB | 252.3 GB | $5.33/h 8x Nvidia A100 |
Qwen1.5 110B
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|
| OriginalQwen/Qwen1.5-110B | vLLM | 222 GB | 252.3 GB | $5.33/h 8x Nvidia A100 |
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").