---

license: other

base_model: Qwen/Qwen2.5-VL-3B-Instruct

base_model_relation: quantized

library_name: gguf

pipeline_tag: text-generation

tags:

• gguf
• llama.cpp
• dashq
• quantized

---

!DASH-Q

Qwen2.5-VL-3B-Instruct - DASH-Q 2-bit GGUF

llama.cpp GGUF files of Qwen/Qwen2.5-VL-3B-Instruct 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 |

| --- | --- | --- | --- |

| Qwen2.5-VL-3B-Instruct-DASHQ-IQ2_XXS.gguf | IQ2_XXS | 0.93 GB | 2.41 |

| Qwen2.5-VL-3B-Instruct-DASHQ-IQ2_XS.gguf | IQ2_XS | 1.06 GB | 2.74 |

| Qwen2.5-VL-3B-Instruct-DASHQ-IQ2_M.gguf | IQ2_M | 1.13 GB | 2.92 |

| Qwen2.5-VL-3B-Instruct-DASHQ-Q2_K_XL.gguf | Q2_K_XL | 1.23 GB | 3.19 |

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) | 0.95 GB | 26.41 | 32.51 |

| IQ2_XXS | unsloth UD-IQ2_XXS | 0.97 GB | 20.61 | 29.39 |

| IQ2_XXS | DASH-Q IQ2_XXS | 0.93 GB | 17.40 | 25.37 |

| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 1.03 GB | 14.89 | 22.79 |

| IQ2_XS | DASH-Q IQ2_XS | 1.06 GB | 13.72 | 20.12 |

| IQ2_M | llama.cpp IQ2_M (imatrix) | 1.14 GB | 12.73 | 19.80 |

| IQ2_M | unsloth UD-IQ2_M | 1.17 GB | 12.53 | 18.84 |

| IQ2_M | DASH-Q IQ2_M | 1.13 GB | 12.60 | 19.00 |

| Q2_K_XL | llama.cpp Q2_K (imatrix) | 1.27 GB | 12.05 | 19.26 |

| Q2_K_XL | unsloth UD-Q2_K_XL | 1.31 GB | 13.03 | 19.25 |

| Q2_K_XL | DASH-Q Q2_K_XL | 1.23 GB | 11.31 | 18.37 |

llama-perplexity, context 2048; WikiText-2 test, C4 validation (256 x 2048 tokens).

Usage

llama-cli -m Qwen2.5-VL-3B-Instruct-DASHQ-Q2_K_XL.gguf -ngl 99 -c 8192

License

Inherits the license of the base model (Qwen/Qwen2.5-VL-3B-Instruct).