Loading…
Loading…
Models / Phi-4-mini
2 models (~3.84B) · ↓ 375.5K 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.
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Originalmicrosoft/Phi-4-mini-instruct | vLLM | 7.7 GB | 11.9 GB | $0.10/h RTX 3060 |
| unsloth bnb 4bit (unsloth)unsloth/Phi-4-mini-instruct-unsloth-bnb-4bit | vLLM | 3.2 GB | 7 GB | $0.10/h RTX 3060 |
| phi 4 mini traffic signal (ebnezr-isaac)ebnezr-isaac/phi-4-mini-traffic-signal | ONNX Runtime GenAI | 2.5 GB | 5.4 GB | $0.08/h GTX 1660 SUPER |
| Version | Runtime | Weights | GPU memory | Cheapest machine now |
|---|---|---|---|---|
| Phi 4 Mini Instruct CPU INT4 ONNX (tonythethompson)tonythethompson/Phi-4-Mini-Instruct-CPU-INT4-ONNX | ONNX Runtime GenAI | 4.9 GB | 8.2 GB | $0.09/h RTX 3060 |
| Phi 4 Mini Instruct GPU INT4 ONNX (tonythethompson)tonythethompson/Phi-4-Mini-Instruct-GPU-INT4-ONNX | ONNX Runtime GenAI | 3.4 GB | 6.5 GB | $0.09/h 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").