Qwen2 models
15 Qwen2 models on Hugging Face, from 494M to 72.7B parameters. At the precision each one is published in, the smallest needs about 1.1 GB of VRAM (Qwen2-0.5B, cheapest live fit: RTX 5060 Ti) and the largest about 163 GB (Qwen2-72B, cheapest live fit: RTX A5000). The cheapest way to run Qwen2-0.5B is $0.110/hr.
Qwen2 models
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| Qwen2-0.5B | 494M | BF16 | 1.1 GB | RTX 5060 Ti | $0.110/hr |
| Qwen2-0.5B-Instruct | 494M | BF16 | 1.1 GB | RTX 5060 Ti | $0.110/hr |
| Qwen2-1.5B | 1.5B | BF16 | 3.5 GB | RTX 5060 Ti | $0.110/hr |
| Qwen2-1.5B-Instruct | 1.5B | BF16 | 3.5 GB | RTX 5060 Ti | $0.110/hr |
| Qwen2-1.5B-Instruct-FP8 | 1.5B | F8_E4M3 | 1.7 GB | RTX 5060 Ti | $0.110/hr |
| gte-Qwen2-1.5B-instruct | 1.8B | F32 | 7.9 GB | V100 | $0.088/hr |
| Qwen2-VL-2B-Instruct | 2.2B | BF16 | 4.9 GB | RTX 5060 Ti | $0.110/hr |
| gte-Qwen2-7B-instruct | 7.6B | F32 | 34.0 GB | RTX A6000 | $0.363/hr |
| Qwen2-7B | 7.6B | BF16 | 17.0 GB | RTX A5000 | $0.176/hr |
| Qwen2-7B-Instruct | 7.6B | BF16 | 17.0 GB | RTX A5000 | $0.176/hr |
| Qwen2-VL-7B-Instruct | 8.3B | BF16 | 18.5 GB | RTX A5000 | $0.176/hr |
| Qwen2-57B-A14B | 57.4B | BF16 | 128 GB | RTX A5000 × 6 | $1.06/hr |
| Qwen2-57B-A14B-Instruct | 57.4B | BF16 | 128 GB | RTX A5000 × 6 | $1.06/hr |
| Qwen2-72B | 72.7B | BF16 | 163 GB | RTX A5000 × 7 | $1.23/hr |
| Qwen2-72B-Instruct | 72.7B | BF16 | 163 GB | RTX A5000 × 7 | $1.23/hr |
VRAM is for the precision the model is published in, with the same overhead and an 8,192-token context assumed on every page; see the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
