Olmo-3-7B-Think - DASH-Q 2-bit GGUF
llama.cpp GGUF files of allenai/Olmo-3-7B-Think 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 |
|---|---|---|---|
Olmo-3-7B-Think-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 2.26 GB | 2.48 |
Olmo-3-7B-Think-DASHQ-IQ2_XS.gguf |
IQ2_XS | 2.58 GB | 2.84 |
Olmo-3-7B-Think-DASHQ-IQ2_M.gguf |
IQ2_M | 2.72 GB | 2.99 |
Olmo-3-7B-Think-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 2.93 GB | 3.22 |
Perplexity (lower is better)
| Type | Model | Size | WikiText-2 | C4 |
|---|---|---|---|---|
| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 2.14 GB | 13.53 | 23.21 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 2.25 GB | 13.27 | 22.67 |
| IQ2_XXS | DASH-Q IQ2_XXS | 2.26 GB | 12.14 | 21.69 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 2.32 GB | 12.31 | 21.50 |
| IQ2_XS | DASH-Q IQ2_XS | 2.58 GB | 10.96 | 19.64 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 2.68 GB | 11.46 | 20.07 |
| IQ2_M | unsloth UD-IQ2_M | 2.72 GB | 11.45 | 19.99 |
| IQ2_M | DASH-Q IQ2_M | 2.72 GB | 10.75 | 19.21 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 2.86 GB | 11.48 | 20.36 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 3.03 GB | 11.34 | 20.02 |
| Q2_K_XL | DASH-Q Q2_K_XL | 2.93 GB | 10.70 | 19.21 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Olmo-3-7B-Think-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (allenai/Olmo-3-7B-Think).