Anthropic launches Claude Fable 5 with strict safety guardrails ahead of 2026 IPO
As Anthropic prepares for a record-breaking market debut alongside SpaceX and OpenAI, the release of Claude Fable 5 marks a significant shift in accessibility, tempered by rigorous safety protocols and a new pricing structure that doubles previous rates.
Anthropic has launched Claude Fable 5, the first publicly accessible iteration of its Mythos-class artificial intelligence model. The release follows the firm’s recent warnings regarding the dangers of rapidly advancing AI systems, positioning the new model as a safer alternative for general users while maintaining strict operational boundaries. The launch coincides with a broader industry shift toward the "MANGOS" corporate hierarchy, as Anthropic, OpenAI, and SpaceX prepare for potentially record-breaking initial public offerings in June 2026.
The model is designed to excel in software engineering, knowledge work, and vision tasks, but it operates under hard safety limits. In high-risk domains such as cybersecurity, biology, chemistry, and distillation, Claude Fable 5 blocks responses and defers to Claude Opus 4.8. This deferral mechanism is intended to mitigate risks associated with the model’s advanced capabilities, which Anthropic has previously described as potentially achieving recursive self-improvement without human intervention.
Access to the model is being rolled out in stages across subscription plans. Until June 22, 2026, Claude Fable 5 is included at no extra cost in Pro, Max, Team, and seat-based Enterprise plans. From June 23, usage will require credits, with the firm stating it intends to restore the model as a standard subscription feature as soon as possible. Concurrently, Anthropic is deploying a separate internal version, Mythos 5, to approved organisations, including those managing critical infrastructure across 15 countries.
To defend against complex and novel attacks, including potential jailbreaks, Anthropic has implemented a mandatory 30-day data retention policy for all traffic. This policy applies even to enterprises that previously held zero-retention agreements. The firm stated that the data will not be used for training but solely to identify and reduce false positives and defend against security threats. This move could set a new industry precedent, linking access to powerful AI models with mandatory data retention as a safety measure.
Pricing for both Claude Fable 5 and Mythos 5 is set at $10 per million input tokens and $50 per million output tokens, which is double the cost of Opus 4.8. The higher price point may serve as a deterrent for some enterprises, particularly those already critical of rising AI costs. However, early adopters such as Rakuten have indicated that the model’s ability to reflect on and validate its own work justifies the expense for highly autonomous operations. Third-party testing by analytics firm Hex and coding platform Base44 has also highlighted the model’s strong performance in complex analytical tasks and application generation.
