Loading…
Loading…
Models / phi-4
1 model (~14.7B) · ↓ 399.9K downloads a month · from $0.18/h on 2x 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 |
|---|---|---|---|---|
| Originalmicrosoft/phi-4 | vLLM | 29 GB | 36.7 GB | $0.45/h Nvidia A40 |
| reasoningmicrosoft/Phi-4-reasoning | vLLM | 29 GB | 36.7 GB | $0.45/h Nvidia A40 |
| reasoning plusmicrosoft/Phi-4-reasoning-plus | vLLM | 29 GB | 36.7 GB | $0.45/h Nvidia A40 |
| FP8 dynamic (RedHatAI)RedHatAI/phi-4-FP8-dynamic | vLLM | 16 GB | 21.5 GB | $0.18/h 2x RTX 3060 |
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").