---

license: llama3.1

base_model: meta-llama/Llama-3.1-8B-Instruct

base_model_relation: quantized

library_name: gguf

pipeline_tag: text-generation

tags:

• gguf
• llama.cpp
• dashq
• quantized

---

!DASH-Q

Llama-3.1-8B-Instruct - DASH-Q 2-bit GGUF

llama.cpp GGUF files of meta-llama/Llama-3.1-8B-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 |

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

| Llama-3.1-8B-Instruct-DASHQ-IQ2_XXS.gguf | IQ2_XXS | 2.51 GB | 2.49 |

| Llama-3.1-8B-Instruct-DASHQ-IQ2_XS.gguf | IQ2_XS | 2.78 GB | 2.77 |

| Llama-3.1-8B-Instruct-DASHQ-IQ2_M.gguf | IQ2_M | 3.01 GB | 2.99 |

| Llama-3.1-8B-Instruct-DASHQ-Q2_K_XL.gguf | Q2_K_XL | 3.28 GB | 3.26 |

Perplexity (lower is better)

| Type | Model | Size | WikiText-2 | C4 |

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

| | F16 | 16.1 GB | 6.40 | 10.90 |

| IQ2_XXS | llama.cpp IQ2_XXS (imatrix) | 2.40 GB | 12.54 | 21.45 |

| IQ2_XXS | unsloth UD-IQ2_XXS | 2.50 GB | 11.78 | 19.48 |

| IQ2_XXS | DASH-Q IQ2_XXS | 2.51 GB | 10.42 | 19.19 |

| IQ2_XS | llama.cpp IQ2_XS (imatrix) | 2.61 GB | 9.97 | 17.29 |

| IQ2_XS | DASH-Q IQ2_XS | 2.78 GB | 8.44 | 15.29 |

| IQ2_M | llama.cpp IQ2_M (imatrix) | 2.95 GB | 8.18 | 14.23 |

| IQ2_M | unsloth UD-IQ2_M | 3.00 GB | 8.00 | 13.80 |

| IQ2_M | DASH-Q IQ2_M | 3.01 GB | 7.72 | 13.60 |

| Q2_K_XL | llama.cpp Q2_K (imatrix) | 3.18 GB | 8.52 | 14.68 |

| Q2_K_XL | unsloth UD-Q2_K_XL | 3.39 GB | 7.89 | 13.55 |

| Q2_K_XL | DASH-Q Q2_K_XL | 3.28 GB | 7.51 | 13.20 |

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

Zero-shot accuracy (lm-eval-harness, 0-shot)

| File | arc_easy | arc_challenge | piqa | hellaswag | winogrande | openbookqa | commonsense_qa | truthfulqa_mc2 | lambada_openai | avg |

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

| DASH-Q IQ2_XXS | 70.0 | 43.5 | 75.4 | 66.7 | 70.4 | 37.8 | 63.1 | 43.9 | 66.7 | 59.72 |

| DASH-Q IQ2_XS | 76.3 | 50.0 | 77.9 | 72.2 | 72.6 | 41.0 | 72.5 | 46.2 | 70.1 | 64.32 |

| DASH-Q IQ2_M | 78.0 | 53.1 | 79.7 | 75.2 | 72.3 | 42.0 | 73.7 | 48.4 | 70.8 | 65.91 |

| DASH-Q Q2_K_XL | 77.3 | 50.2 | 79.6 | 75.6 | 72.5 | 43.4 | 72.8 | 49.4 | 72.5 | 65.92 |

Usage

llama-cli -m Llama-3.1-8B-Instruct-DASHQ-IQ2_M.gguf -ngl 99 -c 8192

License

Inherits the license of the base model (meta-llama/Llama-3.1-8B-Instruct).