Ministral-3-14B-Reasoning-2512 - DASH-Q 2-bit GGUF
llama.cpp GGUF files of mistralai/Ministral-3-14B-Reasoning-2512 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 |
|---|---|---|---|
Ministral-3-14B-Reasoning-2512-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 4.06 GB | 2.40 |
Ministral-3-14B-Reasoning-2512-DASHQ-IQ2_XS.gguf |
IQ2_XS | 4.62 GB | 2.73 |
Ministral-3-14B-Reasoning-2512-DASHQ-IQ2_M.gguf |
IQ2_M | 4.92 GB | 2.90 |
Ministral-3-14B-Reasoning-2512-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 5.37 GB | 3.17 |
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) | 3.92 GB | 7.84 | 13.83 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 4.06 GB | 7.82 | 13.58 |
| IQ2_XXS | DASH-Q IQ2_XXS | 4.06 GB | 7.77 | 13.50 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 4.28 GB | 7.23 | 12.82 |
| IQ2_XS | DASH-Q IQ2_XS | 4.62 GB | 6.87 | 12.24 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 4.84 GB | 6.81 | 12.03 |
| IQ2_M | unsloth UD-IQ2_M | 4.91 GB | 6.75 | 11.89 |
| IQ2_M | DASH-Q IQ2_M | 4.92 GB | 6.61 | 11.75 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 5.25 GB | 6.67 | 11.79 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 5.53 GB | 6.61 | 11.67 |
| Q2_K_XL | DASH-Q Q2_K_XL | 5.37 GB | 6.50 | 11.53 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Ministral-3-14B-Reasoning-2512-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (mistralai/Ministral-3-14B-Reasoning-2512).