Anthropic Preparing New Models Codenamed Mythos and Capybara

Anthropic's Next Generation: Mythos and Capybara Models Leaked

Anthropic appears to be preparing to launch two new AI models codenamed 'Mythos' (v1) and 'Capybara' (v2), according to a leak from a Cloudflare Pages site that surfaced on Hacker News.

What We Know

• Mythos (v1): A new model designation in Claude's lineup
• Capybara (v2): A second new model, possibly a successor or variant
• The leak came from a site at m1astra-mythos.pages.dev
• HN commenters suggest this may have been an accidentally published draft

The Naming Convention

Anthropic's current Claude lineup follows classical naming: Haiku, Sonnet, Opus. The new names continue this:

• Mythos: Greek origin — fitting the classical theme
• Capybara: World's largest rodent — a surprising departure from classical names

Community Speculation

• Mythos could be a 'thinking' or 'reasoning' model similar to OpenAI's o-series
• It may involve multi-agent spawning and extensive verification
• Capybara's unusual name may be an internal codename

Competitive Context

This comes as OpenAI has GPT 5.4, Google advances Gemini, and Claude Opus 4.6 faces competitive pressure. A dual-model launch suggests Anthropic targeting different use cases.