Liontix/ruby-9b-GGUF overview
Warning This model is experimental and not meant for production or serious use. Opus Agent x Gemini 3 Flash Preview This model was trained using SFT Supervised…
Runs locally from ~5.24 GB disk (8 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| Ruby-Q4_K_M.gguf | GGUF | Q4_K_M | 5.24 GB | Download |
Model Details
Model README
---
base_model:
- Liontix/ruby-9b
tags:
- text-generation-inference
- transformers
- unsloth
- qwen3_5
license: apache-2.0
language:
- en
datasets:
- TeichAI/gemini-3-flash-preview
---
Warning
This model is experimental and not meant for production or serious use.
Opus Agent x Gemini 3 Flash Preview
This model was trained using SFT (Supervised Fine Tuning) on Gemini 3 Flash Preview responses, but this time I masked the CoT (Chain of Thought/Thinking Part) when training.
In the future I plan to release more models like these, the catch would be using a fine tuned base model that was SFT with raw/synthetic CoT (not reasoning summaries like before),
and then training on a lot more entries without touching the CoT.
For generating synthetic CoTs I might use Glint-Research/Glint-Trace, seems like a promising model and a step up from just using reasoning summaries.
Thanks to
- armand0e: for the awesome base
- unsloth: for the open source tooling
- Glint-Research: for releasing glint trace (not yet used to create this model) and creating interesting papers on LLMs/NNs Glint-Research-Papers
Metadata
- Developed by: Liontix
- License: apache-2.0
- Finetuned from model : armand0e/Qwen3.5-9B-Opus-Agent
Run Liontix/ruby-9b-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models