squ11z1/Mythos-nano

Hugging Face
TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jun 14, 2026License:mitArchitecture:Transformer0.1K Open Weights Featherless Exclusive Warm

squ11z1/Mythos-nano is a 3 billion parameter independent open model project, not an official Anthropic release. This model excels in competitive mathematics and coding problems, demonstrating strong reasoning capabilities comparable to much larger models. It achieves high scores on benchmarks like AIME and LeetCode contests, making it suitable for complex problem-solving tasks. The model is uncensored, with safety guardrails reduced, and is not recommended for tool-calling or agent-based programming.

Loading preview...

Mythos-nano: A Compact Model for Advanced Reasoning

Mythos-nano is an independent 3 billion parameter language model project, developed by squ11z1, focusing on advanced reasoning in mathematics and coding. Despite its small size, it demonstrates performance comparable to trillion-parameter systems on specific competitive problems.

Key Capabilities & Performance

  • Exceptional Mathematical Reasoning: Mythos-nano achieves high scores on competitive mathematics benchmarks such as AIME25 (91.4%), AIME26 (94.3%), HMMT25 (89.3%), and BruMO25 (93.8%). With CLR (Competitive Learning Rate), its performance further improves, reaching 96.7% on AIME25 and 97.1% on AIME26.
  • Strong Coding Proficiency: The model excels in competitive programming, achieving a 96.1% pass rate on LeetCode contests (Python), placing it near top-tier models like GPT-5.3-Codex and Gemini 3.1 Pro.
  • Uncensored Output: Mythos-nano has reduced safety guardrails, meaning it will not decline requests a safety-tuned model normally would. Users are advised to use it responsibly.

When to Use Mythos-nano

  • Competitive Programming: Ideal for solving LeetCode-style problems and other competitive coding challenges.
  • Complex Mathematical Tasks: Highly effective for problems requiring advanced mathematical reasoning.

Important Considerations

  • Not for Tool-Calling/Agents: This model was not trained on tool-calling or agent-based programming data and is explicitly not recommended for tasks involving function calling, API orchestration, or autonomous coding agents.
  • Responsible Use: Due to its uncensored nature, users are solely responsible for its outputs and legal compliance.

Finetunes
2 models