Claude Fable 5 Goes Public: Anthropic's Mythos-Class AI Reaches General Users With Built-In Safety Routing
News Summary
On June 9, 2026 (Eastern Time), Anthropic made a landmark move in the AI industry by publicly releasing Claude Fable 5, the company's first Mythos-class AI model available to general users. The launch marks the end of a months-long restricted rollout that began in April 2026, when the model was deemed too powerful for open access due to its advanced autonomous capabilities. Simultaneously, Anthropic opened Claude Mythos 5 to users who previously had access through the Mythos Preview program, including enterprise partners under Project Glasswing.
What Is a Mythos-Class Model
Mythos-class is Anthropic's internal classification for models that represent a qualitative leap in reasoning, autonomous problem-solving, and scientific capability. The original Mythos Preview model drew widespread attention because it could autonomously discover thousands of critical and severe cybersecurity vulnerabilities across major operating systems and web browsers — a capability no prior publicly available model had demonstrated at scale. This classification sits above Anthropic's existing Opus tier and signals models capable of operating at or near expert-human level across highly technical domains including software engineering, scientific research, and advanced knowledge work.
Why the Delayed Public Release
When Anthropic first evaluated the Mythos Preview model internally, the company concluded that existing safety infrastructure was insufficient for a broad release. The model's advanced understanding of cybersecurity exploits, biological systems, and chemical processes raised concerns that it could be misused in harmful ways. As a result, access was limited to a small group of vetted institutional partners, and Anthropic's safety teams spent the intervening months building new routing and intervention mechanisms. The transition from a tightly restricted preview to a public product in under three months reflects both the rapid maturation of Anthropic's safety tooling and the competitive pressure the company faces from other frontier labs.
The Safeguard Architecture Behind the Launch
The public release of Claude Fable 5 is made possible by a novel safeguard system that automatically intercepts and re-routes high-risk queries rather than rejecting them outright. When Fable 5 detects requests in sensitive domains — including offensive cybersecurity, biology, chemistry, and model-distillation techniques — it hands the query off to Claude Opus 4.8, a less capable but still highly functional model that does not carry the same risk profile. According to Anthropic's early deployment data, at least 95% of all Fable 5 sessions run entirely on Fable's own responses, meaning the vast majority of legitimate use cases are completely unaffected by the rerouting mechanism.
Performance and Benchmark Results
Claude Fable 5 outperforms competing frontier models from OpenAI and Google DeepMind across all major benchmarks evaluated at launch, according to Anthropic's internal and third-party testing. On select benchmarks, Fable 5 scored more than 10 percentage points higher than Claude Opus 4.8, its immediate predecessor in the Anthropic lineup. The model shows particularly strong results in software engineering tasks, long-horizon knowledge work, and scientific reasoning — areas that make it immediately useful for enterprise customers, researchers, and advanced developers.
Pricing and Availability
Claude Fable 5 is available to enterprise customers and paid subscribers starting June 9, 2026 (Eastern Time). The model is priced at $10 per million input tokens and $50 per million output tokens — approximately twice the cost of Claude Opus 4.8. Anthropic has positioned this pricing as reflective of the model's substantially higher capability ceiling and the infrastructure costs associated with running the safeguard routing system at scale. Claude Mythos 5, the unrestricted version available to prior Mythos Preview participants, is also now live for that cohort.
Broader Industry Context
The release also comes amid reports that Anthropic is planning an IPO, adding a commercial dimension to the timing of the Mythos-class launch. More broadly, the announcement underscores a defining trend in modern AI development: frontier capabilities once considered too sensitive for public access are becoming commercially available as safety engineering catches up with model capability. Anthropic's approach of building a tiered routing system rather than simply withholding the model entirely may serve as a template for how the industry handles future capability thresholds.