Anthropic has officially launched Claude Fable 5, marking the first time its advanced Mythos model is available to the general public. This new iteration boasts significant capabilities in software engineering, knowledge work, and visual processing. However, Anthropic has implemented stringent safety protocols, ensuring that in sensitive domains such as cybersecurity, biology, and chemistry, the model will refrain from responding and instead defer to the Claude Opus 4.8 model.

The Mythos model, initially released as a preview in April, was previously restricted to a select group of partners due to security concerns. Last week, access was broadened to hundreds of organizations across 15 countries, with a particular focus on those managing critical infrastructure. Now, a version of this cutting-edge technology is accessible to all users via Anthropic’s Claude API and its consumption-based Enterprise plans.

Initially, Fable 5 will be included at no additional cost in Pro, Max, Team, and seat-based Enterprise plans until June 22. Following this period, it will transition to a usage-based credit system, with plans to reintegrate it as a standard subscription feature in the near future. Concurrently, Anthropic is rolling out a distinct version, Mythos 5, to organizations that have already received approval for advanced model access.

The release of Fable 5 occurs as Anthropic gears up for a potential public market debut, a move that places it alongside other major AI players like OpenAI and Elon Musk’s SpaceX. This development also follows Anthropic’s recent call for a coordinated pause on the development of frontier AI, citing concerns about the rapid pace of advancement and the potential for recursive self-improvement, where AI systems could autonomously enhance themselves without human intervention.

Anthropic has emphasized its commitment to safety, detailing extensive testing of Fable 5 against potential “jailbreak” attempts. The company conducted both internal and external bug bounty programs, along with red-teaming exercises with external organizations, none of which yielded universal jailbreaks. Despite these efforts, Anthropic acknowledges the possibility of novel attack vectors. To mitigate risks, a mandatory 30-day traffic retention policy will be enforced for all users of Fable 5 and Mythos 5, even for enterprises with prior zero-retention agreements. This data will be used solely for security purposes, such as defending against sophisticated attacks and identifying false positives, and will not be used for model training. This policy may set a new industry standard for accessing powerful AI models.

Key Takeaways

• Anthropic has made its advanced Mythos AI model publicly available through Claude Fable 5, with enhanced capabilities in software engineering, knowledge work, and vision.
• The model incorporates robust safety features, including response blocking in high-risk areas and a mandatory 30-day traffic retention policy for security purposes.
• Fable 5 is priced at $10 per million input tokens and $50 per million output tokens, double the cost of Claude Opus 4.8, potentially influencing adoption rates.

Editor’s Analysis & Impact

The launch of Claude Fable 5 represents a significant step in democratizing access to powerful AI technology, but Anthropic’s cautious approach highlights the ongoing tension between innovation and safety in the AI sector. The stringent safety guardrails and mandatory data retention policy, while framed as necessary precautions, could face scrutiny from users concerned about privacy and operational flexibility. The premium pricing also positions Fable 5 as a high-value tool for enterprises willing to invest in advanced AI capabilities, potentially creating a tiered market for AI services. This move could influence how other AI labs approach public releases of their most advanced models, balancing performance with responsible deployment.

Frequently Asked Questions

Q: What is Claude Fable 5?
A: Claude Fable 5 is the first publicly accessible version of Anthropic's advanced Mythos AI model. It offers enhanced capabilities in areas like software engineering, knowledge work, and vision, but includes safety limitations for high-risk domains.

Q: What are the safety measures implemented in Fable 5?
A: Fable 5 includes hard safety limits that block responses in sensitive areas like cybersecurity and biology, deferring instead to Claude Opus 4.8. Additionally, Anthropic has implemented a mandatory 30-day traffic retention policy for all users to help defend against novel attacks and jailbreaks.

Q: How is Fable 5 priced?
A: Fable 5 is priced at $10 per million input tokens and $50 per million output tokens. This is double the price of Anthropic's Claude Opus 4.8 model.