Can I run Phi-4 14B locally?
Yes. At Q4 with 8K context, Phi-4 14B needs about 11 GB, so it runs entirely on any GPU with 12 GB of VRAM or more, starting with the AMD Radeon RX 7700 XT.
Memory needed by compression and context length
| Compression | 4K context | 8K context | 16K context |
|---|---|---|---|
Q4 The most common download. Small quality loss. | 10 GB | 11 GB | 12 GB |
Q5 Very close to full quality. | 11 GB | 12 GB | 14 GB |
Q8 Practically indistinguishable from FP16 at about half the size. | 16 GB | 17 GB | 19 GB |
FP16 Uncompressed. Reference quality, largest size. | 29 GB | 30 GB | 31 GB |
Total memory: model weights, context (KV cache) and about 1 GB of runtime overhead. To run entirely on the GPU, VRAM must be about 5% larger than this.
GPUs that run Phi-4 14B entirely on the GPU
- Amazon
AMD Radeon RX 7700 XT
12 GB · ~32 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 3060
12 GB · ~27 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 3060 12GB
12 GB · ~27 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 2060 12GB
12 GB · ~25 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 4070 Ti
12 GB · ~37 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 4070 SUPER
12 GB · ~37 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 5070
12 GB · ~49 tok/s at Q4
- Amazon
NVIDIA GeForce RTX 4070
12 GB · ~37 tok/s at Q4
- Amazon
AMD Radeon RX 6750 XT
12 GB · ~32 tok/s at Q4
- Amazon
AMD Radeon RX 7700
12 GB · — at Q4
Check it with your hardware
Change the compression, context length and system RAM in the calculator.
Open calculator