Mythos-class capability is now accessible to any enterprise on Anthropic's API. The hard safety limits and mandatory data retention policy that come with it will set the terms for how the industry handles frontier model access going forward.
Key Takeaways
- Claude Fable 5 is Anthropic’s first Mythos-class model available to the general public, via API and consumption-based Enterprise plans
- Hard blocks are in place for cybersecurity, biology, chemistry, and distillation — falling back to Opus 4.8 in those areas
- All traffic now carries a mandatory 30-day retention policy, even for enterprises with prior zero-retention agreements
What Happened
Anthropic launched Claude Fable 5 today, the first publicly available version of its Mythos-class model. The model is now accessible through Anthropic’s Claude API and consumption-based Enterprise plans, with a staged subscription rollout running through June 22 at no extra cost on Pro, Max, Team, and seat-based Enterprise plans.
Fable 5 is built for software engineering, knowledge work, and vision tasks, but ships with hard safety limits. In high-risk areas including cybersecurity, biology, chemistry, and distillation, the model blocks responses entirely and falls back to Claude Opus 4.8. Early data shows at least 95% of Fable sessions run entirely on the model’s own responses without triggering a fallback.
Anthropic is simultaneously deploying Mythos 5 to organizations already approved under its restricted access program. Mythos launched as a limited preview in April, then expanded last week to hundreds of organizations across 15 countries focused on critical infrastructure.
Pricing for both models is set at $10 per million input tokens and $50 per million output tokens, double the cost of Opus 4.8. On June 23, Fable 5 will be removed from subscription plans and require usage credits until Anthropic can restore it as a standard feature.
Why It Matters
The mandatory 30-day data retention policy is the detail most enterprises will push back on. Anthropic says it will not use retained traffic for model training, only to defend against novel jailbreak attempts and reduce false positives. But for organizations that negotiated zero-retention agreements specifically to protect sensitive workflows, the policy removes a protection they built compliance assumptions around. This launch also follows Anthropic’s public call for major AI labs to establish a coordinated brake on frontier development, warning that systems may soon reach recursive self-improvement, autonomously improving without human intervention.
The skeptic case is straightforward: Anthropic spent months warning that Mythos-class models carry risks serious enough to restrict access to vetted critical infrastructure operators. Now the same capability is available to any developer with an API key and a credit balance. The safety argument for public release rests entirely on the classifiers holding, and classifiers have a documented history of failing under sustained adversarial pressure. Anthropic’s own disclosure — that novel attacks remain possible despite 1,000+ hours of red-teaming — is the honest version of that caveat.
At the highest effort, Fable reflects on and validates its own work. For us, that’s what makes highly autonomous operations possible — the extra thinking pays for itself. — Rakuten
Bottom Line
Watch the June 23 transition closely. Pulling Fable 5 from subscriptions and moving it to usage credits mid-rollout is an unusual move, and how Anthropic handles pricing transparency in that window will signal how it plans to manage frontier model access as it heads toward its IPO.
For operators building on Anthropic’s API, the 30-day retention mandate needs a compliance review before any Fable 5 workloads go into production. The capability jump is real and the benchmark scores from Hex, Genspark, and Base44 are credible signals worth testing, tracked by Relve, an AI tools intelligence platform.
