Phi-4-reasoning - DASH-Q 2-bit GGUF
llama.cpp GGUF files of microsoft/Phi-4-reasoning quantized with DASH-Q, at several 2-bit-class sizes. Every file uses only standard llama.cpp tensor types (no tensor above 4 bits) and loads in any recent llama.cpp build.
| File | Type | Size | Bits / weight |
|---|---|---|---|
Phi-4-reasoning-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 4.23 GB | 2.30 |
Phi-4-reasoning-DASHQ-IQ2_XS.gguf |
IQ2_XS | 4.87 GB | 2.65 |
Phi-4-reasoning-DASHQ-IQ2_M.gguf |
IQ2_M | 5.20 GB | 2.83 |
Phi-4-reasoning-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 5.69 GB | 3.10 |
Perplexity (lower is better)
| Type | Model | Size | WikiText-2 | C4 |
|---|---|---|---|---|
| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 4.55 GB | 8.71 | 16.23 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 4.27 GB | 11.20 | 20.49 |
| IQ2_XXS | DASH-Q IQ2_XXS | 4.23 GB | 9.57 | 18.41 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 4.92 GB | 7.92 | 14.97 |
| IQ2_XS | DASH-Q IQ2_XS | 4.87 GB | 7.98 | 15.67 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 5.49 GB | 7.23 | 13.91 |
| IQ2_M | unsloth UD-IQ2_M | 5.26 GB | 8.16 | 15.45 |
| IQ2_M | DASH-Q IQ2_M | 5.20 GB | 7.42 | 14.49 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 5.92 GB | 7.29 | 13.97 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 5.80 GB | 7.64 | 14.66 |
| Q2_K_XL | DASH-Q Q2_K_XL | 5.69 GB | 7.23 | 14.12 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Phi-4-reasoning-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (microsoft/Phi-4-reasoning).