arnomatic/gpt-oss-20b-heretic-scanner-V1-2

Model Overview

This model, arnomatic/gpt-oss-20b-heretic-scanner-V1-2, is a 20 billion parameter large language model with a 32,768 token context length. It is a decensored variant of OpenAI's gpt-oss-20b, created using the Heretic v1.1.0 tool. The primary modification involves a significant reduction in refusal rates, dropping from 98 out of 100 in the original model to 12 out of 100 in this version, as measured by KL divergence analysis.

Key Capabilities & Differentiators

• Decensored Output: Achieves a substantially lower refusal rate compared to the base gpt-oss-20b model, making it suitable for use cases requiring less restrictive content generation.
• OpenAI's gpt-oss Foundation: Inherits the core architecture and capabilities of the gpt-oss series, designed for powerful reasoning and agentic tasks.
• Configurable Reasoning: Supports adjustable reasoning effort (low, medium, high) via system prompts, allowing users to balance speed and detail.
• Agentic Features: Includes native capabilities for function calling, web browsing, and Python code execution.
• Fine-tunable: The 20B parameter size makes it suitable for fine-tuning on consumer hardware.
• Permissive Licensing: Based on a model released under the Apache 2.0 license, enabling broad commercial and experimental deployment.

Good For

• Applications requiring a less restrictive content policy than the original gpt-oss-20b.
• Developer use cases involving agentic tasks, function calling, and code execution where content filtering is not desired.
• Experimentation and customization through fine-tuning on accessible hardware.