Hugging Face Overview
Model Overview
DavidAU/Gemma-3-1B-it-GLM-4.7-Flash-Heretic-Uncensored-Thinking is a 1 billion parameter Gemma-based instruction-tuned model, developed by DavidAU. It is distinguished by its fully uncensored and deep reasoning capabilities, achieved through fine-tuning with the GLM 4.7 reasoning dataset via Unsloth. The model is designed to provide direct, detailed, and compact reasoning without typical refusal behaviors.
Key Capabilities
- Uncensored Output: Engineered to generate content without refusals, even for sensitive or explicit requests, though it may require explicit directives for graphic or slang content.
- Deep Reasoning: Incorporates a "deep thinking" mechanism, which can be activated explicitly or automatically, leading to detailed and precise outputs.
- 32k Context Length: Supports extended conversational and analytical tasks.
- Temperature Stability: Reasoning remains stable across a wide temperature range (0.1 to 2.5).
- Performance: Achieves competitive benchmarks for its size, including 0.344 on arc_challenge and 0.720 on piqa.
- Low Refusal Rate: Demonstrates significantly reduced refusals (3/100) compared to the original Gemma model (99/100), with a low KL divergence of 0.33.
Good For
- Applications requiring unfiltered and direct responses.
- Use cases where deep, explicit reasoning is paramount.
- Scenarios demanding creative or sensitive content generation without built-in censorship.
- Developers seeking a compact model with high reasoning stability.