Fable 5 and Mythos 5 Back Online: Anthropic Restores Model Access

Anthropic returned two of its most capable models to users on July 1, 2026. The release notes for Claude applications that day state, briefly but unambiguously: access is being restored to both Claude Fable 5 and Claude Mythos 5. This closes a nearly three-week period during which both models were unavailable to everyone — from individual subscribers to enterprise teams.

The return was announced the day before, on June 30, in a post titled “Redeploying Fable 5,” stating that Fable 5 would return globally on July 1. What happened today is not a new announcement but the actual execution of that plan — the models are live again.

The Models Are Live Again

Fable 5 is of Mythos class and Anthropic positions it above the Opus line in capability; it carries a one-million-token context window and a new tokenizer. Mythos 5 is the other model that was affected by the same suspension. For many users who spent the last three weeks working with alternative models, the key message is simple: both models are available again with no special conditions stated in the release notes.

The release note itself is deliberately sparse and points to the broader announcement on Anthropic’s website for background on the decision.

Why Were Fable 5 and Mythos 5 Withdrawn in the First Place?

The root of the problem was geopolitical, not technical. On June 12, the US government imposed export controls requiring that access to Fable 5 and Mythos 5 be restricted exclusively to US citizens. Since Anthropic cannot verify user citizenship in real time, the only compliant option was to suspend both models for everyone — rather than a selective shutdown.

Behind the scenes was also a concrete security finding: Amazon researchers had discovered a technique by which Fable 5, given specific prompting, could be induced to identify software vulnerabilities and demonstrate their exploitation. That finding formed part of the case for the restriction.

What Changed Before the Return

Two conditions had to be met before the return. The first was regulatory: the government lifted the export controls on June 30, removing the legal barrier to global availability. Mythos 5 received the green light for restored access somewhat earlier, on June 26.

The second condition was safety-related. Anthropic trained an improved safety classifier targeted precisely at the technique discovered by Amazon researchers; according to the company, the classifier blocks that pattern in over 99 percent of cases. Also important was an argument that weakened the case for unique danger: it emerged that less capable models can also replicate the same vulnerability identification, meaning Fable 5 did not offer a unique “uplift” that would make it disproportionately risky.

A Framework for Rating Jailbreak Severity

Alongside the model’s return, Anthropic also proposed an industry framework for rating jailbreak severity — a shared standard for ranking how dangerous any given safety bypass actually is. The proposal is being developed together with Amazon, Microsoft, Google, and other Glasswing partners, with the aim of measuring incidents like the one that triggered the suspension consistently and comparably across the industry.

What This Means for Users

For end users the practical message is straightforward: Fable 5 and Mythos 5 are back in the lineup as of today, and the entire episode shows how intertwined frontier model availability and the regulatory framework have become. The model launched on June 9, withdrawn on June 12, and returned on July 1 completed in three weeks the full cycle of launch, geopolitical hold, and technical remediation — and came back with a tangibly strengthened safety layer.