Anthropic has officially introduced a public version of its highly anticipated next-generation AI system, giving users their first opportunity to interact with technology that had previously been reserved for limited testing. However, the release comes with important restrictions, and the version now available is not identical to the more unrestricted model that generated considerable attention within the AI community.
The newly released model, called Claude Fable 5, has been integrated into the Claude desktop application. According to Anthropic, the system represents a major step forward in AI capabilities and has the potential to deliver meaningful benefits across a wide range of industries and use cases.
At the same time, the company acknowledges that the model’s advanced cybersecurity knowledge introduces potential risks. Anthropic warns that the technology could be misused if left unchecked, which is why Fable 5 includes stricter safeguards designed to prevent harmful behavior. These protections may occasionally block legitimate requests, although the company estimates such overly cautious responses should occur in only a small percentage of interactions.
Alongside Fable 5, Anthropic is also introducing a separate version known as Claude Mythos 5. Both systems share the same underlying architecture, but Mythos 5 operates with fewer restrictions in certain cybersecurity-related areas. Access to that version remains limited to a select group of cybersecurity professionals and critical infrastructure organizations.
Performance data released by Anthropic highlights significant gains in several areas, including software development, reasoning involving physical and spatial concepts, and overall problem-solving ability. However, the most notable improvements appear in cybersecurity tasks, which have been a major focus of discussion ever since reports of the Mythos project first surfaced.
Subscribers to Claude’s paid plans can begin using Fable 5 immediately. Access extends across multiple subscription tiers, allowing a broad range of paying customers to experiment with the new model. The availability, however, is only temporary under the current rollout strategy.
Anthropic has confirmed that Fable 5 will be removed from standard subscription access after June 23. Once that happens, users who wish to continue using the model will need to purchase usage credits, with costs calculated similarly to API-based access. The company says this decision is tied to infrastructure capacity and resource management concerns.
The firm has stated that it intends to restore Fable 5 as a regular subscription feature in the future, but only after sufficient computing resources become available to support broader access.
Users testing the model should also expect significantly higher resource consumption compared to previous flagship Claude models. As a result, usage limits may be reached much faster during extended sessions, particularly when tackling complex tasks.
Early experiences with the new model have not been entirely seamless. Some users reported encountering unexpected errors shortly after launching Fable 5-powered tools, including vague messages indicating technical issues with the model. Whether these problems are temporary launch-day hiccups or signs of heavier demand remains unclear.
Despite the restrictions and occasional technical challenges, the release of Fable 5 marks an important milestone for Anthropic. It offers a glimpse into a new generation of AI systems that are both more capable and more controversial, balancing extraordinary potential with the growing need for safeguards and responsible deployment.