```html

June 9, 2026 Anthropic released Claude Fable 5 — the first publicly available model of the Mythos class. But the press release hides a detail that changes the entire picture: Fable 5 and Mythos 5 are one and the same model. The only difference is which safeguards are activated.

What "public version" really means

Fable 5 is the first commercially available model of the Mythos class. According to Anthropic, the new model's capabilities surpass all previously publicly available models from the company. However, the difference between the two models lies exclusively in safeguards — Anthropic chose separate names precisely to make this distinction clear.

Mythos 5 — the same base model with safeguards disabled in certain areas — is available only to verified organizations in cybersecurity and infrastructure through the Glasswing project. Cyber defenders and critical infrastructure providers gained access to Glasswing; a separate program for biomedical researchers is planned for the coming weeks.

"In April, we launched Project Glasswing, releasing the first Mythos-class model only to a limited group of cyber defenders. We hope to eventually make Mythos-level capabilities available to everyone — provided we develop sufficiently robust safeguards."

Anthropic, official announcement

Technical parameters: what actually changed

Both models share a common foundation: a context window of 1 million tokens, up to 128,000 output tokens per request, support for images, tools, memory, and computer use.

Claude Fable 5 includes safety classifiers that can reject certain requests. Separate classifiers monitor conversations and detect abuse attempts in three areas: cybersecurity, biology, and chemistry. According to Anthropic, the safeguards are intentionally configured conservatively — some legitimate requests are also blocked.

The price of entry — and the hidden cost of refusals

Fable 5 and Mythos 5 cost $10 per million input tokens and $50 per million output tokens. But there's a nuance in pricing for refusals: if a harmful request is redirected to Opus 4.8 instead of Fable 5, the user pays at Opus rates. If the request is blocked mid-conversation — the initial tokens are charged at Fable rates, and subsequent ones at Opus rates.

A new distribution logic: not "who is stronger," but "who is allowed"

Anthropic argues that Mythos-class models have crossed a threshold where they pose significant risks — without safeguards, Fable 5's capabilities in cybersecurity could be used to cause serious harm.

The division of one product into public and closed versions establishes a new reality: the most powerful public models are now not simply "the best available" — they are the best permitted. Between Fable 5 and full-fledged Mythos 5 stands not technology, but the company's decision about who can be trusted.

If Fable 5's conservative safeguards block legitimate scientific or research requests more often than Anthropic acknowledges publicly — competitors without such restrictions will gain a structural advantage precisely in the segment that Fable 5 targets.

```