Anthropic made its most capable AI model available to the general public for the first time on Tuesday, launching Claude Fable 5 — a version of its previously restricted Mythos model — through the company’s API and enterprise plans. The move brings frontier AI capabilities to a broader audience, but with a set of hard safety guardrails that block responses in high-risk domains.
Fable 5 excels at software engineering, complex knowledge work, and vision-based tasks, according to Anthropic. However, in areas the company considers high-risk — including cybersecurity, biology, chemistry, and model distillation — the system automatically defers to a less powerful fallback model, Claude Opus 4.8, instead of generating its own response.
Also read: The 'Tokenpocalypse' is here: Microsoft's GitHub Copilot price hike signals a broader AI reckoning
From Restricted Partner Access to Public Availability
The underlying technology, Mythos, first launched as a preview in April 2026 to a small group of partners. Anthropic cited cybersecurity concerns as the reason for the narrow initial release. Last week, the company expanded access to hundreds of organizations across 15 countries, specifically those managing critical infrastructure. Now, a version of that same technology is available to any developer or enterprise through Anthropic’s Claude API and consumption-based Enterprise plans.
Access for subscription users is being rolled out in stages. Through June 22, Fable 5 is included in Pro, Max, Team, and seat-based Enterprise plans at no extra cost. On June 23, Anthropic will remove the model from those subscription tiers, requiring usage credits going forward. The company says it plans to restore Fable 5 as a standard subscription feature as soon as possible.
Also read: Reid Hoffman leaves Microsoft board to go 'founder mode' with AI drug startup Manus
Anthropic also released Mythos 5, an updated version of its advanced model, to organizations that were already approved to access the earlier Mythos system.
Safety Testing and Data Retention
Before releasing Fable 5, Anthropic said it stress-tested the model’s safety classifiers against jailbreak attempts. The company reported that an internal bug bounty program produced no universal jailbreaks in over 1,000 hours of testing, and external red-teaming organizations also failed to find universal attacks. Anthropic acknowledged, however, that novel attacks may still be possible.
As a result, the company is implementing a new policy: all traffic to Fable 5 and Mythos 5 will be retained for 30 days, even for enterprises that previously had zero-retention agreements. Anthropic said it will not use the retained data for training, but only to defend against complex and novel attacks and to reduce false positives in its safety classifiers. The policy could set an industry precedent, tying access to increasingly powerful models to mandatory data retention framed as a safety measure.
For most queries, Fable 5 will handle responses directly. Anthropic said early data shows that at least 95% of Fable sessions run entirely on the model’s own responses, with the fallback to Opus 4.8 triggered only in rare cases.
Pricing and Performance
Pricing for both Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens — double the price of Opus 4.8. That cost alone may deter widespread use, particularly as many enterprises are already growing critical of AI spending after seeing bills exceed annual budgets.
In third-party testing, analytics company Hex said Fable was the first model to score 90% on its core analytics benchmark of complex, long-running analytical tasks. Vibe-coding platform Base44 noted in a statement that Fable is better at building full applications in a single attempt and has excellent tool-calling capabilities. AI-powered workspace platform Genspark said Fable outperformed every other model in its evaluations, particularly on UI design and game coding tasks.
Shopping rewards platform Rakuten, an early tester, highlighted the model’s ability to validate its own work. “At the highest effort, Fable reflects on and validates its own work,” Rakuten said in a statement. “For us, that’s what makes highly autonomous operations possible — the extra thinking pays for itself.”
Broader Context: IPOs and Frontier AI Risks
Fable’s public launch comes as Anthropic prepares to enter public markets, alongside OpenAI and Elon Musk’s SpaceX. It also follows a recent plea from Anthropic urging major global AI labs to establish a coordinated safety mechanism — what the company calls a “brake pedal” — on frontier AI development. Anthropic has warned that systems are advancing so rapidly that they may soon achieve recursive self-improvement (RSI), autonomously improving themselves without human intervention.
Anthropic said it expects demand for Fable 5 to be very high and difficult to predict. For enterprises willing to pay the premium, the model offers capabilities that were previously locked behind partnership agreements — but with safety policies that may reshape how the industry balances access and risk.
