Anthropic has released Claude Fable 5, its most capable AI model yet for general use, while simultaneously offering a more permissive Claude Mythos 5 to a limited group of trusted partners. The launch marks a major step in the company’s effort to make frontier-level AI broadly available without opening the door to misuse in high-risk areas such as cybersecurity, biology, and chemistry.
Claude Fable 5 Launch
Anthropic announced on June 8, 2026 that Claude Fable 5 is now generally available and that it is the first Mythos-class model made safe for public use. The company says Fable 5 is its most capable model ever released to the public, with performance improvements aimed at coding, knowledge work, analysis, research, and long-context tasks.
The model is available through Anthropic’s own platform and through cloud and enterprise distribution channels, including AWS. CNBC and other outlets reported that Anthropic is positioning Fable 5 as the broad-access version of the company’s most advanced model family, rather than restricting it to a small set of vetted users.
Claude Mythos 5 and Restricted Access
Alongside Fable 5, Anthropic introduced Claude Mythos 5, a less restricted version intended for approved organizations participating in trusted-access programs such as Project Glasswing. The company says Mythos 5 is the same underlying model family, but with fewer safety restrictions for vetted users, including cybersecurity defenders and critical infrastructure partners.
That distinction matters because Anthropic had previously kept Mythos limited over concerns that its capabilities could be misused. Reuters, CNBC, and Anthropic’s own announcement all point to the same basic structure: Fable 5 for wider public use, Mythos 5 for a narrower trusted cohort.
Safety Guardrails and Fallbacks
Anthropic says the broad release of Fable 5 was made possible by a new safety system that routes certain high-risk prompts to a less capable model instead of answering directly. The company says prompts involving cybersecurity, biology, chemistry, and model distillation can be diverted to Claude Opus 4.8, which is used as a fallback for sensitive requests.
According to Anthropic, this layered setup is meant to reduce the risk of harmful outputs while preserving most of Fable 5’s capabilities for normal users. The company also says its safety classifiers were tested extensively and designed to resist jailbreak attempts. In Anthropic’s framing, this is a way to make a very powerful model broadly available without fully exposing users to the highest-risk behaviors of the Mythos tier.
Pricing and Enterprise Access
Anthropic and reporting from CNBC say Fable 5 and Mythos 5 are priced at 10 dollars per million input tokens and 50 dollars per million output tokens. That puts the models at roughly twice the cost of Anthropic’s previous flagship public model, Claude Opus 4.8.
The pricing signals that Anthropic is targeting serious enterprise, developer, and research use cases rather than casual consumer experimentation. AWS also confirmed availability of Fable 5 for customers through Bedrock and Claude on AWS, with standard enterprise controls and guardrails. This makes the new model a major product for businesses that want advanced AI with compliance and infrastructure support.
Why Anthropic Is Moving Carefully
Anthropic’s rollout comes after the company reportedly restricted Mythos access in the past because of cybersecurity concerns. That history explains why the company is now emphasizing safety architecture as much as raw capability.
The move also reflects a broader industry pattern: the most powerful models are increasingly being released with usage limits, fallback logic, and vetted-access programs rather than as fully open systems. Anthropic’s strategy appears designed to strike a balance between leadership in model performance and caution around misuse.
Competition in the AI Market
The launch puts Anthropic back into direct competition with OpenAI and other frontier AI labs at a time when model quality, safety, and enterprise trust are all central to market positioning. CNBC described the rollout as a significant milestone in the race among AI developers to deliver increasingly capable systems while addressing safety concerns.
The release also shows how Anthropic is differentiating itself: rather than simply scaling up power, it is wrapping its most advanced model class in a set of controls meant to make it publicly deployable. That approach could matter a lot for enterprise buyers deciding which model family to build into products and internal systems.
What the Release Signals
The biggest takeaway is that Anthropic is not just adding another model to its lineup; it is formalizing a two-track structure for frontier AI access. One track is broad but constrained, and the other is less restricted but limited to trusted users.
That structure may become a template for the next phase of AI commercialization, especially if regulators and enterprise customers continue to demand stronger controls around dangerous capabilities. For now, Anthropic is betting that safety engineering can make a model powerful enough for cutting-edge work while still acceptable for broad use.
