Mistral-Small-3.1-24B-Instruct-2503 - DASH-Q 2-bit GGUF
llama.cpp GGUF files of mistralai/Mistral-Small-3.1-24B-Instruct-2503 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 |
|---|---|---|---|
Mistral-Small-3.1-24B-Instruct-2503-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 6.76 GB | 2.29 |
Mistral-Small-3.1-24B-Instruct-2503-DASHQ-IQ2_XS.gguf |
IQ2_XS | 7.68 GB | 2.60 |
Mistral-Small-3.1-24B-Instruct-2503-DASHQ-IQ2_M.gguf |
IQ2_M | 8.25 GB | 2.80 |
Mistral-Small-3.1-24B-Instruct-2503-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 9.15 GB | 3.10 |
Text only: the vision tower is not included.
Perplexity (lower is better)
| Type | Model | Size | WikiText-2 | C4 |
|---|---|---|---|---|
| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 6.55 GB | 22.77 | 46.24 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 6.75 GB | 22.87 | 44.04 |
| IQ2_XXS | DASH-Q IQ2_XXS | 6.76 GB | 7.19 | 12.20 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 7.21 GB | 16.62 | 29.68 |
| IQ2_XS | DASH-Q IQ2_XS | 7.68 GB | 6.17 | 10.52 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 8.11 GB | 9.17 | 15.32 |
| IQ2_M | unsloth UD-IQ2_M | 8.24 GB | 8.96 | 14.79 |
| IQ2_M | DASH-Q IQ2_M | 8.25 GB | 5.74 | 9.86 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 8.89 GB | 6.61 | 10.64 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 9.29 GB | 6.36 | 10.21 |
| Q2_K_XL | DASH-Q Q2_K_XL | 9.15 GB | 5.58 | 9.56 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Mistral-Small-3.1-24B-Instruct-2503-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (mistralai/Mistral-Small-3.1-24B-Instruct-2503).