---

language:

• en

license: mit

library_name: gguf

tags:

• text-generation
• dflash
• speculative-decoding
• draft-model
• block-diffusion
• llama.cpp
• quantized
• reasoning
• ornith

base_model: ornith-ai/Ornith-1.5-9B-GGUF

base_model_relation: quantized

pipeline_tag: text-generation

---

Ornith-1.5-9B DFlash GGUF

Official GGUF quantizations of ornith-ai/Ornith-1.5-9B-DFlash, the speculative decoding draft model designed to accelerate ornith-ai/Ornith-1.5-9B inference in llama.cpp.

Overview

This is a speculative decoding draft model (Block-Diffusion / DFlash architecture), not a standalone language model.

It runs alongside Ornith-1.5-9B GGUF to increase generation throughput by drafting up to 16 tokens per step (block_size = 16).

!Ornith-1.5 w/ Different Decoding Acceleration Techniques

GGUF Files

| File | Quantization | Size |

| :--- | :---: | :---: |

| Ornith-1.5-9B-DFlash-Q4_K_M.gguf | Q4_K_M (Recommended) | 730.48 MiB |

| Ornith-1.5-9B-DFlash-Q6_K.gguf | Q6_K | 1021.29 MiB |

| Ornith-1.5-9B-DFlash-Q8_0.gguf | Q8_0 | 1319.66 MiB |

| Ornith-1.5-9B-DFlash-BF16.gguf | BF16 | 2474.66 MiB |

Usage

Pair this draft model with your target Ornith-1.5-9B model in llama.cpp using the -md (model draft) flag:

llama-cli \
  -m Ornith-1.5-9B-Q4_K_M.gguf \
  -md Ornith-1.5-9B-DFlash-Q4_K_M.gguf \
  --spec-type draft-dflash \
  -ngl 99 -c 4096 --temp 0.6 \
  -p "<|im_start|>user\nWrite a quicksort in Python.<|im_end|>\n<|im_start|>assistant\n"

Attribution & License

• Distributed under the MIT License, matching upstream ornith-ai/Ornith-1.5-9B-DFlash.
• Target Model: ornith-ai/Ornith-1.5-9B.
• Reference GGUF release: ornith-ai/Ornith-1.5-9B-GGUF.