Phi 4 Mini Instruct needs about 4.1 GB of memory at Q4_K_M (a 2.5 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 23 quants, from 1.5 GB (IQ2_M) to 4.1 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Phi 4 Mini Instruct is a 3.84B-parameter Phi model in GGUF format, quantized by lucasxie2025. Its trained context length is 128K tokens. The repository has 23 quant files, from 1.5 GB (IQ2_M) to 4.1 GB (Q8_0).
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| IQ2_M | 1.5 GB | 3.1 GB | Fits | Get file → | |
| IQ3_XXS | 1.7 GB | 3.3 GB | Fits | Get file → | |
| Q2_K | 1.7 GB | 3.3 GB | Fits | Get file → | |
| Q2_K_L | 1.8 GB | 3.4 GB | Fits | Get file → | |
| IQ3_XS | 1.8 GB | 3.4 GB | Fits | Get file → | |
| Q3_K_S | 1.9 GB | 3.5 GB | Fits | Get file → | |
| IQ3_M | 2.0 GB | 3.6 GB | Fits | Get file → | |
| Q3_K_M | 2.1 GB | 3.7 GB | Fits | Get file → | |
| IQ4_XS | 2.2 GB | 3.8 GB | Fits | Get file → | |
| Q3_K_L | 2.2 GB | 3.8 GB | Fits | Get file → | |
| IQ4_NL | 2.3 GB | 3.9 GB | Fits | Get file → | |
| Q4_0 | 2.3 GB | 3.9 GB | Fits | Get file → | |
| Q4_K_S | 2.3 GB | 3.9 GB | Fits | Get file → | |
| Q3_K_XL | 2.4 GB | 4.0 GB | Fits | Get file → | |
| Q4_K_M | 2.5 GB | 4.1 GB | Fits | Get file → | |
| Q4_1 | 2.5 GB | 4.1 GB | Fits | Get file → | |
| Q4_K_L | 2.6 GB | 4.2 GB | Fits | Get file → | |
| Q5_K_S | 2.7 GB | 4.3 GB | Fits | Get file → | |
| Q5_K_M | 2.8 GB | 4.4 GB | Fits | Get file → | |
| Q5_K_L | 3.0 GB | 4.6 GB | Fits | Get file → | |
| Q6_K | 3.2 GB | 4.7 GB | Fits | Get file → | |
| Q6_K_L | 3.3 GB | 4.9 GB | Fits | Get file → | |
| Q8_0 | 4.1 GB | 5.7 GB | Fits | Get file → |
What GPU do you need for Phi 4 Mini Instruct?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q3_K_XL (2.4 GB file, about 4.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (4.1 GB file, about 5.7 GB in use).
Which Phi 4 Mini Instruct quant should you choose?
Q4_K_M is the usual starting point for Phi 4 Mini Instruct: it balances file size and quality well. Here it is a 2.5 GB file.
The Q8_0 file is 4.1 GB, about 1.6 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is IQ2_M at 1.5 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Phi 4 Mini Instruct
How much VRAM does Phi 4 Mini Instruct need?
About 4.1 GB at Q4_K_M with an 8K context window. That is the 2.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Phi 4 Mini Instruct on an 8 GB GPU?
Yes, with Q8_0: the 4.1 GB file needs about 5.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Phi 4 Mini Instruct GGUF?
It depends on the quant: IQ2_M is 1.5 GB, IQ3_XXS is 1.7 GB, Q2_K is 1.7 GB, Q2_K_L is 1.8 GB, IQ3_XS is 1.8 GB, Q3_K_S is 1.9 GB and 17 more.
What context length does Phi 4 Mini Instruct support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download Phi 4 Mini Instruct GGUF?
The files are on Hugging Face in the lucasxie2025/microsoft_Phi-4-mini-instruct-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Phi 4 Mini Instruct use?
Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.