DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic
DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic is a 27 billion parameter Qwen3.6-based model fine-tuned by DavidAU. This instruction-tuned model focuses on enhancing core intelligence, demonstrating improved performance across various benchmarks compared to its base Qwen3.6-27B-Instruct and Qwen3.5-27B-Instruct counterparts. It is optimized for general intelligence tasks, making it suitable for applications requiring robust reasoning capabilities.
Loading preview...
Model Overview
DavidAU/Qwen3.6-27B-F451-AND-TRI-Polar-Ultra-Pro-Writer-Uncensored-Heretic is a 27 billion parameter language model, fine-tuned by DavidAU using Unsloth on local hardware. This model is specifically engineered to elevate its core intelligence, building upon the Qwen3.6 architecture.
Key Capabilities & Enhancements
This fine-tuned version demonstrates significant improvements over its base models, Qwen3.6-27B-Instruct and Qwen3.5-27B-Instruct, particularly in core intelligence benchmarks. The model was tested in "Instruct" mode, which generally yields better results with the testing harness.
- Enhanced Core Intelligence: The primary goal of this fine-tune was to boost the model's fundamental reasoning and understanding capabilities.
- Benchmark Performance: It shows higher scores across various benchmarks (arc, arc/e, boolq, hswag, obkqa, piqa, wino) compared to the non-heretic base models. For instance, its mxfp8 scores are 0.665 for arc and 0.830 for arc/e, surpassing the Qwen3.6-27B-Instruct's 0.637 and 0.798 respectively.
- Instruction-Tuned: Optimized for instruction-following tasks, making it responsive and effective in conversational or command-based scenarios.
Use Cases
This model is well-suited for applications requiring a strong foundation in general intelligence and robust instruction following. Its improved benchmark performance suggests suitability for tasks demanding accurate reasoning and comprehensive understanding.