Qwen3.5-122B-A10B - DASH-Q 2-bit GGUF
llama.cpp GGUF files of Qwen/Qwen3.5-122B-A10B 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 |
|---|---|---|---|
Qwen3.5-122B-A10B-DASHQ-IQ2_XXS.gguf |
IQ2_XXS | 35.92 GB | 2.31 |
Qwen3.5-122B-A10B-DASHQ-IQ2_XS.gguf |
IQ2_XS | 38.34 GB | 2.47 |
Qwen3.5-122B-A10B-DASHQ-Q2_K_XL.gguf |
Q2_K_XL | 41.05 GB | 2.64 |
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) | 34.04 GB | 6.30 | 11.26 |
| IQ2_XXS | unsloth UD-IQ2_XXS | 36.64 GB | 4.89 | 9.84 |
| IQ2_XXS | DASH-Q IQ2_XXS | 35.92 GB | 4.98 | 9.99 |
| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 37.68 GB | 5.63 | 10.60 |
| IQ2_XS | DASH-Q IQ2_XS | 38.34 GB | 4.81 | 9.82 |
| IQ2_M | llama.cpp IQ2_M (imatrix) | 41.69 GB | 5.21 | 10.16 |
| IQ2_M | unsloth UD-IQ2_M | 39.15 GB | 4.82 | 9.77 |
| Q2_K_XL | llama.cpp Q2_K (imatrix) | 46.32 GB | 5.25 | 10.20 |
| Q2_K_XL | unsloth UD-Q2_K_XL | 41.85 GB | 4.71 | 9.65 |
| Q2_K_XL | DASH-Q Q2_K_XL | 42.47 GB | 4.75 | 9.72 |
llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).
Usage
llama-cli -m Qwen3.5-122B-A10B-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192
License
Inherits the license of the base model (Qwen/Qwen3.5-122B-A10B).