Qwen3.5-0.8B - DASH-Q 2-bit GGUF
llama.cpp GGUF files of Qwen/Qwen3.5-0.8B 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 |
|---|---|---|---|
Qwen3.5-0.8B-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 0.31 GB | 3.15 |
Qwen3.5-0.8B-DASHQ-IQ2_XS.gguf |
IQ2_XS | 0.32 GB | 3.34 |
Qwen3.5-0.8B-DASHQ-IQ2_M.gguf |
IQ2_M | 0.34 GB | 3.45 |
Qwen3.5-0.8B-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 0.35 GB | 3.58 |
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) | 0.36 GB | 67.07 | 99.05 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 0.34 GB | 52.29 | 71.67 |
| IQ2_XXS | DASH-Q IQ2_XXS | 0.31 GB | 24.33 | 35.55 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 0.38 GB | 38.48 | 54.90 |
| IQ2_XS | DASH-Q IQ2_XS | 0.32 GB | 20.45 | 28.73 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 0.39 GB | 27.22 | 37.00 |
| IQ2_M | unsloth UD-IQ2_M | 0.37 GB | 23.76 | 33.06 |
| IQ2_M | DASH-Q IQ2_M | 0.34 GB | 19.14 | 26.78 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 0.43 GB | 26.48 | 38.65 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 0.42 GB | 21.79 | 30.64 |
| Q2_K_XL | DASH-Q Q2_K_XL | 0.35 GB | 18.65 | 26.28 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Qwen3.5-0.8B-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (Qwen/Qwen3.5-0.8B).