---

license: llama3.1

base_model: allenai/Llama-3.1-Tulu-3.1-8B

tags:

- gguf

- llama.cpp

- allenai

- tulu

- text-generation

---

Llama-3.1-Tulu-3.1-8B GGUF

GGUF conversion of allenai/Llama-3.1-Tulu-3.1-8B, published by Mu2 Solutions.

Converted with llama.cpp build b9859 (4fc4ec554) using the official Hugging Face to GGUF converter, then quantized with llama-quantize. This is the first GGUF release of this model.

License

llama3.1 (Meta Llama Community License) — this model is derived from Llama 3.1 weights via Allen Institute for AI's Tulu 3.1 training recipe. The license of the base model applies. See the source model card for details.

Files

| File | Quant | Size |

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

| Llama-3.1-Tulu-3.1-8B-Q4_K_M.gguf | Q4_K_M | 4.6 GB |

| Llama-3.1-Tulu-3.1-8B-Q8_0.gguf | Q8_0 | 8.0 GB |

Q4_K_M is the recommended default for most use. Q8_0 for near-lossless quality with more RAM.

Quickstart

# llama-server (OpenAI-compatible)
llama-server -m Llama-3.1-Tulu-3.1-8B-Q4_K_M.gguf -c 8192 -ngl 99

# one-shot completion
llama-completion -m Llama-3.1-Tulu-3.1-8B-Q4_K_M.gguf -p "Your prompt here" -n 256

# interactive chat
llama-cli -m Llama-3.1-Tulu-3.1-8B-Q4_K_M.gguf -c 8192 -ngl 99

Verification

• Conversion: HF safetensors (4 shards) → F16 GGUF → quantized Q4_K_M / Q8_0
• Validated: single-turn completion produces correct, coherent output with the Tulu chat template (<|user|> / <|assistant|>)

Credit

• Model: Allen Institute for AI (Ai2)
• Conversion: Mu2 Solutions — promoting truly open-source AI, one GGUF at a time