granite-4.1-3b - DASH-Q 2-bit GGUF
llama.cpp GGUF files of ibm-granite/granite-4.1-3b 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 |
|---|---|---|---|
granite-4.1-3b-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 1.05 GB | 2.46 |
granite-4.1-3b-DASHQ-IQ2_XS.gguf |
IQ2_XS | 1.18 GB | 2.78 |
granite-4.1-3b-DASHQ-IQ2_M.gguf |
IQ2_M | 1.25 GB | 2.95 |
granite-4.1-3b-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 1.35 GB | 3.18 |
Perplexity (lower is better)
| Type | Model | Size | WikiText-2 | C4 |
|---|---|---|---|---|
| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 1.02 GB | 16.62 | 26.62 |
| IQ2_XXS | DASH-Q IQ2_XXS | 1.05 GB | 11.97 | 19.81 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 1.11 GB | 13.41 | 20.88 |
| IQ2_XS | DASH-Q IQ2_XS | 1.18 GB | 10.79 | 17.05 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 1.24 GB | 11.33 | 17.89 |
| IQ2_M | unsloth UD-IQ2_M | 1.29 GB | 10.78 | 17.09 |
| IQ2_M | DASH-Q IQ2_M | 1.25 GB | 10.13 | 16.12 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 1.37 GB | 10.65 | 17.00 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 1.41 GB | 10.13 | 16.16 |
| Q2_K_XL | DASH-Q Q2_K_XL | 1.35 GB | 10.06 | 15.75 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m granite-4.1-3b-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (ibm-granite/granite-4.1-3b).