Qwen3.5-4B - DASH-Q 2-bit GGUF
llama.cpp GGUF files of Qwen/Qwen3.5-4B 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 |
|---|---|---|---|
Qwen3.5-4B-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 1.44 GB | 2.66 |
Qwen3.5-4B-DASHQ-IQ2_XS.gguf |
IQ2_XS | 1.57 GB | 2.90 |
Qwen3.5-4B-DASHQ-IQ2_M.gguf |
IQ2_M | 1.65 GB | 3.04 |
Qwen3.5-4B-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 1.75 GB | 3.22 |
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) | 1.61 GB | 14.21 | 21.39 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 1.52 GB | 13.28 | 19.75 |
| IQ2_XXS | DASH-Q IQ2_XXS | 1.44 GB | 11.31 | 17.10 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 1.70 GB | 11.96 | 18.09 |
| IQ2_XS | DASH-Q IQ2_XS | 1.57 GB | 10.16 | 15.14 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 1.81 GB | 10.11 | 15.72 |
| IQ2_M | unsloth UD-IQ2_M | 1.76 GB | 10.38 | 15.12 |
| IQ2_M | DASH-Q IQ2_M | 1.65 GB | 9.62 | 14.31 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 1.98 GB | 10.36 | 15.80 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 1.94 GB | 10.12 | 14.83 |
| Q2_K_XL | DASH-Q Q2_K_XL | 1.75 GB | 9.59 | 14.18 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Qwen3.5-4B-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (Qwen/Qwen3.5-4B).