A new large language model (LLM) called Omnicoder-Claude-4.6-Opus-Uncensored-GGUF has been released, designed for local inference and derived from Claude Opus.

Key Features

  • Architecture: Based on Qwen 3.5 9B.
  • Distillation: Created through distillation from Claude Opus.
  • Uncensored: Designed to operate without filters.
  • Quantization: Q4_K_M and Q8_0 quantizations are available.
  • Merge: The model is the result of a merge process that includes models from Jackrong, HauhauCS, and Tesslate.

Technical Details

The model was created using an Add Difference merge script, preserving the GGUF header and metadata structure to ensure compatibility. The creation involved several pre-existing models, including:

  • The latest update of the Jackrong model, trained on a dataset distilled from Claude Opus.
  • The HauhauCS uncensored Qwen 3.5 9B model.
  • Omnicoder created by Tesslate.
  • Bartowski quantization was used as a base.

For those evaluating on-premise deployments, there are trade-offs to consider. AI-RADAR offers analytical frameworks on /llm-onpremise for evaluating these options.