richardyoung/mythos-9b-merged-heretic

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jun 26, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The richardyoung/mythos-9b-merged-heretic is an 8 billion parameter uncensored fine-tuned agent model, based on Qwen3-9B, developed by Richard Young. It is a decensored version of King3Djbl/mythos-9b-merged, created using Heretic v1.4.0, and features a native thinking mode for complex multi-step tasks. This model is optimized for tool use, shell commands, and reasoning, demonstrating significantly reduced refusals compared to its original counterpart. It is best suited for general agent applications requiring robust tool calls and advanced reasoning capabilities.

Loading preview...

Model Overview

The richardyoung/mythos-9b-merged-heretic is an 8 billion parameter uncensored agent model, derived from the Qwen3-9B architecture. Developed by Richard Young, this model is a decensored variant of King3Djbl/mythos-9b-merged, processed with Heretic v1.4.0. It is specifically fine-tuned with 47,824 agent traces, encompassing tool use, shell commands, and reasoning tasks, and incorporates a native thinking mode for handling complex multi-step operations.

Key Differentiators & Performance

This model stands out due to its significantly enhanced censorship resistance, exhibiting only 7 refusals out of 100, a substantial reduction from the original model's 79 refusals. Performance metrics highlight its strengths:

  • Censorship Resistance: Achieves 4.5/5, providing full answers across 9 out of 10 categories.
  • Tool-Use: Scores 4.8/5, proficient in shell, code, Docker, Kubernetes, and SQL.
  • Reasoning: Rated 4.5/5 for logic, debugging, and system design.

Training and License

The model was fine-tuned on the FableForge Mix A dataset, which includes agent traces, shell commands, code generation, and multi-step reasoning examples. It is released under the Apache 2.0 License.

Ideal Use Cases

  • General Agent Applications: Excels in scenarios requiring autonomous agent behavior.
  • Tool Calling: Highly effective for tasks involving shell commands, code execution, and interaction with various tools.
  • Complex Reasoning: Suitable for applications demanding logical deduction, debugging, and system design capabilities.

Finetunes
1 models