Qwen3.6-35B-A3B - DASH-Q 2-bit GGUF
llama.cpp GGUF files of Qwen/Qwen3.6-35B-A3B 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 |
|---|---|---|---|
Qwen3.6-35B-A3B-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 10.53 GB | 2.37 |
Qwen3.6-35B-A3B-DASHQ-IQ2_XS.gguf |
IQ2_XS | 11.20 GB | 2.52 |
Qwen3.6-35B-A3B-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 12.03 GB | 2.71 |
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) | 9.97 GB | 8.16 | 12.84 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 10.76 GB | 6.44 | 10.96 |
| IQ2_XXS | DASH-Q IQ2_XXS | 10.53 GB | 6.53 | 11.07 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 10.98 GB | 7.43 | 12.03 |
| IQ2_XS | DASH-Q IQ2_XS | 11.20 GB | 6.38 | 10.85 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 12.14 GB | 6.99 | 11.32 |
| IQ2_M | unsloth UD-IQ2_M | 11.52 GB | 6.37 | 10.87 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 13.42 GB | 6.70 | 11.24 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 12.29 GB | 6.29 | 10.70 |
| Q2_K_XL | DASH-Q Q2_K_XL | 12.50 GB | 6.29 | 10.74 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Qwen3.6-35B-A3B-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (Qwen/Qwen3.6-35B-A3B).