---

license: apache-2.0

base_model: FINAL-Bench/Darwin-27B-RSI

library_name: gguf

pipeline_tag: text-generation

tags:

• gguf
• llama.cpp
• q4_k_m
• darwin
• darwin-rsi
• recursive-self-improvement
• rsi
• no-human-labels
• reasoning
• thinking
• decision-index
• jev
• vidraft
• final-bench

---

<p align="center">

<a href="https://vidraft.net"><img src="https://img.shields.io/badge/🌐_VIDRAFT-vidraft.net-111111?style=for-the-badge" alt="VIDRAFT"></a>

<a href="https://huggingface.co/FINAL-Bench/Darwin-27B-RSI"><img src="https://img.shields.io/badge/🧬_BF16-Darwin--27B--RSI-blue?style=for-the-badge" alt="BF16"></a>

</p>

Darwin-27B-RSI-GGUF

Q4_K_M GGUF of FINAL-Bench/Darwin-27B-RSI — Darwin-27B-Opus improved by Recursive Self-Improvement, with zero human-written answers.

| File | Quant | Size |

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

| Darwin-27B-RSI-Q4_K_M.gguf | Q4_K_M | 16.8 GB |

Internal checks show no GPQA Diamond accuracy difference between this Q4_K_M file and BF16. It was used as the reasoning path of Darwin-27B-JEV on the Decision Index, and lets the whole engine fit on one 96 GB GPU.

Run

llama-server -m Darwin-27B-RSI-Q4_K_M.gguf --jinja -ngl 99 -c 65536 --port 7931

Thinking model: allow a generous token budget; read the answer after the reasoning block.

Not affiliated with TypeSafe AI or its Jev product. Developer: VIDRAFT · FINAL-Bench.