Phi-3 Medium 14B Instruct

Running Phi-3 Medium 14B Instruct locally

At Q4_K_M and 4K of context, Phi-3 Medium 14B Instruct needs about 9.7 GB — 8.1 GB of weights, 0.78 GB of KV cache and a 0.9 GB activation buffer. The smallest card in this index that clears that comfortably is the RTX 5070 at 12 GB, and 55 of the 63 cards here do. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.

What longer context costs

Going from 4K to 32K adds about 5.47 GB, taking the total to 15.7 GB. Weights do not move with context — only the KV cache does, and it grows linearly, so this is the number to watch when planning for long documents. The model's native window is 128K; holding all of it at Q4_K_M would need about 36 GB.

Which build to download

This index tracks 2 formats for it — GGUF, AWQ — across 3 levels. Q6_K carries the lowest published perplexity loss at 0.8%. The fastest level measured here is AWQ INT4 at 135 tok/s on an RTX 4090, batch 1. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 and CUDA only.

Common questions

How much VRAM does Phi-3 Medium 14B Instruct need?

About 9.7 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 15.7 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.

Will Phi-3 Medium 14B Instruct run on a 12GB GPU?

Yes — at Q4_K_M and 4K context it needs about 9.7 GB, which leaves 2.3 GB spare on a RTX 5070. That is the smallest card in this index that clears it comfortably; 55 of 63 do.

Which quantization of Phi-3 Medium 14B Instruct should I use?

Q6_K has the lowest published quality loss (0.8%), and Q4_K_M is the level most people run. All 3 levels in the index are Q4_K_M, Q6_K, AWQ INT4.