Phi-2 models
1 Phi-2 model from Microsoft on Hugging Face, from 2.8B to 2.8B parameters, published by Microsoft in F16, with a 2K-token context. The smallest official model, phi-2, needs about 6.2 GB of VRAM at its published precision; the cheapest live fit is V100 at $0.088/hr.
Part of the Phi series · Previous generation: Phi-1 · Next generation: Phi-3
Pick a size
One row per official Phi-2 size: the VRAM it needs at each precision, the cheapest GPU that holds it today, and the KV cache for a 32K-token context.
| Model | Parameters | Native VRAM | FP8 VRAM | INT4 VRAM | Live GPU fit (native) | Est. $/hr | KV cache at 32K |
|---|---|---|---|---|---|---|---|
| phi-2 | 2.8B | 6.2 GB | 3.1 GB | 1.6 GB | V100 | $0.088/hr | 10.0 GB |
VRAM is weights times a flat 1.2 overhead; the KV cache is a separate per-model figure at 16-bit, shown where the architecture is published. See the methodology.
Official models (1)
VRAM is for the precision the model is published in: the weights times a flat 1.2 overhead, with the KV cache not included. 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.
