Anthropic Launches Claude Fable 5 and Mythos 5 With New AI Safety Controls

Anthropic has announced the launch of Claude Fable 5, describing it as the most powerful artificial intelligence model the company has made widely available to users. The release marks a significant milestone for Anthropic, as Fable 5 becomes the first broadly accessible model from the company’s highly capable Mythos family of AI systems.

According to Anthropic, Claude Fable 5 delivers strong performance across a range of demanding tasks, including software engineering, knowledge work, and vision-related applications. The company said the model’s advantages become increasingly apparent as tasks grow longer, more complex, and require sustained reasoning.

First Broad Release From the Mythos Family

The launch is notable because Fable 5 is derived from Anthropic’s Mythos class of models, which the company had previously described as too capable in certain areas, particularly cybersecurity, to be released publicly without additional safeguards.

To address those concerns, Anthropic developed new safety mechanisms designed to restrict responses in specific high-risk domains. The company said these systems made it possible to release Fable 5 to a broader audience while maintaining tighter controls over sensitive capabilities.

When the model encounters requests that fall within restricted categories, Fable 5 automatically blocks the response and transfers the interaction to Claude Opus 4.8. Anthropic identified cybersecurity and biology as two areas where these safeguards may be activated.

Despite the restrictions, the company reported that the vast majority of interactions proceed without interruption. During testing, Anthropic said 95% of Fable 5 sessions were completed entirely using the model itself, without requiring a fallback to Opus 4.8.

Anthropic Introduces Claude Mythos 5

Alongside Fable 5, Anthropic is also releasing Claude Mythos 5. However, the company provided relatively few details about the new model.

Anthropic said Mythos 5 is built on the same underlying technology as Fable 5 but operates with some safeguards removed or relaxed in certain areas. As a result, access to Mythos 5 will remain limited initially.

The model is currently available only to organizations that already have access to Claude Mythos Preview, a program managed through Anthropic’s private Project Glasswing initiative. Existing Mythos Preview users will be able to upgrade directly to Mythos 5.

Anthropic said it intends to broaden availability over time through a more structured trusted-access program, though the company did not provide a timeline for wider deployment.

Premium Pricing for Advanced Models

The new models come with significantly higher pricing than Anthropic’s existing Claude Opus 4.8 offering.

Both Claude Fable 5 and Claude Mythos 5 are priced at:

• $10 per million input tokens
• $50 per million output tokens

Anthropic noted that these rates are double the cost of Claude Opus 4.8 but remain half the price of the earlier Mythos Preview offering.

Continued Focus on AI Safety

The release of Fable 5 and Mythos 5 reflects Anthropic’s ongoing effort to balance rapid advances in AI capabilities with safety considerations. The company has repeatedly emphasized the potential risks associated with increasingly powerful AI systems and has argued that stronger safeguards may be necessary as technology continues to evolve.

Anthropic has also stated that future AI development could eventually require a global pause if capabilities advance faster than society’s ability to manage associated risks.

For now, the company is moving forward with broader access to its most advanced models while maintaining restrictions around the capabilities it considers most sensitive.