Custom 3-bit coding model optimized for local agent workflows on 16 GB GPUs. Features a 128K context window for large codebases, long-running tasks, and coding assistants such as OpenCode. Designed for efficient local inference with strong code generation