sillykiwi/Gemma3-4B-CodeCenturion
Gemma3-4B-CodeCenturion is a 4 billion parameter model created by sillykiwi, based on the Gemma 3 architecture. This model is a merge of vanta-research/scout-4b and GetSoloTech/Gemma3-Code-Reasoning-4B, using coder3101/gemma-3-4b-it-heretic as its base. It is specifically designed for programming and cyber/information-security tasks, combining strong coding abilities with cooperative problem-solving and reduced hallucinations.
Loading preview...
Overview
Gemma3-4B-CodeCenturion is a 4 billion parameter language model developed by sillykiwi, built upon the Gemma 3 architecture. This model is a merge of three distinct base models, leveraging their individual strengths to create a specialized tool. It was constructed using mergekit-gui to combine vanta-research/scout-4b, GetSoloTech/Gemma3-Code-Reasoning-4B, and coder3101/gemma-3-4b-it-heretic.
Key Capabilities
- Enhanced Cooperative Problem Solving: Integrates the personality and cooperative problem-solving abilities from
vanta-research/scout-4b. - Strong Code Generation: Incorporates the robust coding capabilities of
GetSoloTech/Gemma3-Code-Reasoning-4B. - Reduced Hallucinations: Utilizes
coder3101/gemma-3-4b-it-hereticas a base, which is noted for its abliteration technique to reduce hallucinations, particularly when addressing complex queries.
Good For
This model is primarily intended for applications in:
- Programming: Excels in code-related tasks due to its specialized training.
- Cybersecurity: Designed to be useful for various cyber and information security-related challenges.
Quantized versions (GGUF and iMatrix-GGUF) are available courtesy of team mradermacher.