camel911/Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETIC

TEXT GENERATIONPricing:Input $1.2 / Output $4.8Concurrent Unit Cost:1Model Size:12.2BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Oct 2, 2026Architecture:Transformer Featherless Exclusive Cold

The camel911/Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETIC is a 12.2 billion parameter Mistral Nemo Instruct model, fine-tuned with Claude Opus 4.5, Gemini, and GPT5.2 high-reasoning datasets. It features an uncensored 'HERETIC' version with significantly reduced refusal rates (14/100) and enhanced 'thinking/reasoning' capabilities, optimized for detailed, complex, and high-quality output generation. This model is designed for use cases requiring unfiltered responses and advanced reasoning across a 32768 token context length.

Loading preview...

Model Overview

The camel911/Mistral-Nemo-2407-12B-Thinking-Claude-Gemini-GPT5.2-Uncensored-HERETIC is a 12.2 billion parameter language model built upon the Mistral Nemo Instruct architecture. This version has undergone a unique two-stage fine-tuning process. First, it was "Heretic'ed" to be fully uncensored, drastically reducing refusal rates from 87/100 to 14/100. Subsequently, it was fine-tuned with three high-reasoning datasets from Claude Opus 4.5, Gemini, and GPT5.2, transforming it into a "thinking/reasoning" model.

Key Capabilities

  • Enhanced Reasoning: Integrates advanced reasoning capabilities from multiple high-performing models, leading to more detailed, complex, and higher-quality outputs.
  • Uncensored Output: Provides unfiltered responses across a wide range of topics, including NSFW content, vivid prose, and intense details, without typical LLM guardrails.
  • Compact Thinking Engine: Generates reasoning blocks that are typically 3-6 paragraphs (300-600 tokens) in size, improving performance without being overly verbose.
  • Flexible Temperature Settings: Reasoning activation is not affected by temperature, allowing for a wide range of creative outputs (0.1 to 2.5 or higher).
  • Extended Context: Supports a context length of 32768 tokens, with suggestions for optimal performance at 8k+.

Good For

  • Creative Writing & Roleplay: Excels in generating vivid, detailed, and uncensored narratives, particularly for horror, intense scenarios, and complex character interactions.
  • Unfiltered Content Generation: Ideal for applications requiring direct, unconstrained responses without built-in censorship or nanny-like behavior.
  • Complex Problem Solving: Benefits from its integrated reasoning engine for tasks demanding structured thought processes and detailed explanations.
  • Exploratory & Experimental Use Cases: Suitable for users who need a model that adheres strictly to user prompts without imposing ethical or content restrictions.

Finetunes
1 models