Amrmostafa25/Mythos-weights
VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 2, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold
Qwythos-9B, developed by Empero, is a 9 billion parameter reasoning model built on a Qwen3.5-9B base, fine-tuned on over 500 million tokens of Claude Mythos and Fable traces. It features a 1,048,576-token context window, native function calling, and self-correction capabilities. This model excels in cybersecurity, biomedical, and quantitative reasoning tasks, offering enhanced performance over its base model.
Loading preview...
Qwythos-9B: A Specialized Reasoning Model
Qwythos-9B, developed by Empero, is a 9 billion parameter model built upon a Qwen3.5-9B base, specifically fine-tuned for advanced reasoning. It leverages over 500 million tokens from Claude Mythos and Fable traces, incorporating chain-of-thought generated by Empero AI's internal rethink tool.
Key Capabilities
- 1,048,576-token context window: Achieved via YaRN rope-scaling, enabling whole-codebase reasoning, multi-document research, and long agentic trajectories.
- Enhanced Reasoning Performance: Demonstrates significant improvements over the base Qwen3.5-9B, with +34 pts on MMLU and +30 pts on GSM8K-strict.
- Native Function Calling: Supports OpenAI/Qwen3.5-style function calling out-of-the-box, without requiring additional wrappers or fine-tuning.
- Self-Correction with Tools: Successfully uses Python executors and web search tools to provide source-cited, factually correct answers across complex domains like math, cybersecurity, and clinical pharmacology.
- Uncensored Design: Intentionally uncensored to engage substantively with technically demanding questions in specialized fields, avoiding boilerplate disclaimers.
Good For
- Cybersecurity and Red-Teaming: Detailed walkthroughs and explanations of complex methodologies.
- Biomedical and Pharmacology: In-depth reasoning on biological mechanisms, drug interactions, and clinical medicine.
- Quantitative Reasoning: Strong performance on multi-step word problems and competition math.
- Agentic Workflows: Deployment-ready for retrieval-augmented agentic settings where factual verification is critical.