Analog of the potentially dangerous AI Mythos: Anthropic releases powerful Claude Fable 5 model to the general public

Anthropic, the developer of the Claude chatbot, has announced the release of its new flagship model, Claude Fable 5. This is the first Mythos-class model to become available not only to a limited circle of select companies but also to a wide range of enterprise clients and paying subscribers.

From limited access to public launch

Just two months ago, Anthropic very cautiously introduced Mythos to select partners under the Project Glasswing initiative. The reason for the limited release was concern that the model could be used to carry out sophisticated cyberattacks. The company now states that it has implemented sufficiently reliable safety mechanisms and is ready to open access to a model of comparable power.

Claude Fable 5 demonstrates "exceptional performance" in software engineering, code analysis, and complex intellectual tasks. On several benchmarks, it surpasses the previous Claude Opus 4.8 model by more than 10%.

Enhanced safety measures

To prevent misuse, additional protective barriers have been implemented in the new model. When attempting to ask a question from a "high-risk" zone (for example, how to create biological weapons or carry out a cyberattack), the model blocks the response and switches to the less powerful Claude Opus 4.8 version.

At the same time, Anthropic has released an updated version of Claude Mythos 5 — the same base model, but with partially lifted restrictions for trusted users and partners.

Strategic timing

The release of the powerful model comes as Anthropic actively prepares for a stock market listing. The company recently filed a confidential application with the SEC. In May, its annualized revenue reached $47 billion (up from $10 billion the previous year), and its valuation grew to $965 billion, surpassing OpenAI.

Claude Fable 5's pricing is significantly higher than previous models: $10 per million input tokens and $50 per million output tokens. However, according to company representatives, users report substantially higher return on investment due to improved quality and efficiency.

In brief

Anthropic has officially released Claude Fable 5 — a Mythos-level model available to a wide audience. Two months ago, the company was hesitant to open access to such power due to risks of misuse. Now, with enhanced safety mechanisms, the model is coming to market. The launch coincides with Anthropic's preparation for an IPO and reflects intense competition in the segment of the most powerful AI systems.