empero-ai/Qwythos-9B-Claude-Mythos-5-1M
Qwythos-9B-Claude-Mythos-5-1M by Empero is a 9 billion parameter reasoning model built on a Qwen3.5-9B base, post-trained on over 500 million tokens of Claude Mythos and Fable traces. It features a 1,048,576-token context window via YaRN rope-scaling and native function calling for tool use. This model excels at complex reasoning tasks across cybersecurity, biomedical, and quantitative domains, demonstrating significant improvements over its base in MMLU and GSM8K benchmarks.
Loading preview...
Qwythos-9B: A Powerful 9B Reasoning Model
Qwythos-9B, developed by Empero, is a 9 billion parameter reasoning model built upon a deeply uncensored Qwen3.5-9B base. It has been extensively post-trained on over 500 million tokens of high-quality Claude Mythos and Fable traces, with in-house generated chain-of-thought data, resulting in a compact and highly capable model.
Key Capabilities & Differentiators
- 1,048,576-token Context Window: Features YaRN rope-scaling enabled by default, providing a 1M-token context window suitable for whole-codebase reasoning, multi-document research, and long agentic trajectories.
- Enhanced Reasoning Performance: Demonstrates significant improvements over its base model, with +34 pts on MMLU and +30 pts on gsm8k-strict under matched evaluation conditions.
- Native Function Calling: Supports OpenAI/Qwen3.5-style function calling out-of-the-box, enabling seamless integration with tools without extra wrappers or fine-tuning.
- Self-Correction with Tools: Proven to self-correct and produce source-cited, factually correct answers on challenging prompts when paired with Python executor and web search tools.
- Uncensored Design: Intentionally uncensored to engage seriously with technically demanding questions in domains like cybersecurity, red-teaming, biology, pharmacology, and clinical medicine, where other models might refuse or hedge.
Ideal Use Cases
Qwythos-9B is particularly well-suited for:
- Complex Reasoning: Excels in cybersecurity, biomedical analysis, and quantitative problem-solving.
- Agentic Workflows: Its long context and tool-use capabilities make it deployment-ready for retrieval-augmented agentic settings.
- Long-form Content Analysis: Capable of handling extensive documents and codebases for tasks like architectural review or multi-paper synthesis.
It's important to note that Qwythos is a reasoning model, and its outputs typically include a <think> block before the final answer. Recommended sampling settings (temperature=0.6, top_p=0.95, top_k=20, repetition_penalty=1.05) are crucial for optimal performance and to avoid repetition loops.