Heidelberg AICurriculum

Claude Mythos is Finally Here.

Nate Herk | AI Automation ·2026-06-09 ·2 min read

Summary written by us from the video's transcript. The video, and everything in it, is Nate Herk | AI Automation's work.

Learn about Claude Fable 5 and Mythos 5 models, their availability, pricing, capabilities, and how they differ in safety safeguards.

Takeaways

  • Fable 5 is free for all subscription tiers only until June 22; after that it will cost $10/$50 per million tokens.
  • Mythos 5 is a restricted, cyber‑secure version of the same model, available through Project Glass Wing partners.
  • Both models are priced at twice Opus’s rate but deliver markedly better performance across coding, knowledge work, vision, and cybersecurity benchmarks.
  • Update your Claude integration now to access Fable 5 and experiment with longer agentic loops before the free window ends.
  • Future public releases may incorporate Mythos‑level capabilities into Opus once safety measures are solidified.

Model Availability and Pricing

Claude Fable 5 is currently available to all users and included for free in Pro, Max, Team, and seat‑based enterprise plans until June 22. After that date it will require token‑based usage credits at $10 per million input tokens and $50 per million output tokens – double the cost of Opus.

Claude Mythos 5 is restricted to Glasswing partners and a small group of cyber defenders; it uses the same underlying model as Fable 5 but with cybersecurity safeguards enabled.

Understanding the Mythos Class

Mythos‑class models sit one tier above Opur in capability. The first release, Claude Mythos preview, launched via Project Glass Wing and was five times more expensive than Opus. The new Mythos 5 costs only twice as much as Opus, indicating it’s a less‑restricted version of the full model.

Fable 5 is marketed as an “extended” Opus for complex, long‑running work, while Mythos 5 adds strong cybersecurity features and is deployed through Project Glass Wing in collaboration with the US government.

Performance Benchmarks

Both Fable 5 and Mythos 5 outperform previous Claude models on benchmarks covering software engineering, knowledge work, vision, scientific research, legal reasoning, biology, and especially cybersecurity.

The blog shows large jumps in agentic coding (SWE‑Bench), multi‑disciplinary reasoning, and complex analytical tasks compared to Opus 4.8 and GPT‑5.5.

Practical Usage Tips

Update your Claude codebase to see the Fable 5 option; it appears under “Meet Fable 5, our newest model for complex, long‑running work.”

Take advantage of the free access window (now until June 22) to experiment with longer agentic loops and vision tasks such as video editing, slide‑deck creation, and website verification.

Safety and Future Outlook

Mythos 5’s cybersecurity safeguards are kept for a limited partner group to mitigate misuse; the same model without those guards is offered as Fable 5.

Anthropic may gradually bake Mythos‑specific capabilities into future Opus releases once safety guardrails are deemed sufficient, meaning broader public access could arrive later under different branding.