Anthropic announced the official release of two new artificial intelligence models, Claude Fable 5.2 and Mythos 5.2, positioning them among the most capable systems in the company's history. The frontier AI lab stated that the latest software iterations bring substantial performance improvements to coding, reasoning, and vision tasks. Most notably, the company claimed that these new models successfully addressed two previously unsolved physics problems concerning black holes, designed working electronic devices, and produced a computer-verifiable version of the proof for Fermat’s Last Theorem.
Key takeaways
- Anthropic released Claude Fable 5.2 and Mythos 5.2 with significant upgrades to coding, reasoning, and vision.
- Fable 5.2 coordinated over 1,000 subagents to build a functional prototype operating system from scratch in eight days.
- Both models are priced at $8 per million input tokens and $40 per million output tokens, generating outputs up to 30% faster than previous versions.
- Mythos 5.2 remains restricted to approved organizations participating in the Cyber Verification Program and Life Sciences Verification Program.
Performance metrics and economic efficiency
Fable 5.2 becomes the company's highest-performing publicly available model for open-ended technical work, outperforming prior iterations across multiple agentic capability evaluations and coding benchmarks. Anthropic also stated that the new models outperform competing frontier models on several evaluations. Alongside these performance gains, the company managed to reduce operational costs and increase processing speed. Outputs generate up to 30% faster than the previous Fable 5.1 and Mythos 5.1 models were able to produce.
Pricing for both new models has dropped to $8 per million input tokens and $40 per million output tokens. This new pricing structure makes them slightly cheaper than OpenAI’s most powerful model, GPT-6 Astra. Fable 5.2 rolls out across Anthropic’s platforms starting immediately, featuring increased usage limits for paid subscribers.
Complex problem solving and engineering milestones
To demonstrate the practical capabilities of the new systems, Anthropic shared several complex test results. The company tasked Fable 5.2 with building a prototype operating system entirely from scratch. The AI model successfully coordinated more than 1,000 subagents and wrote the code for a functioning operating system in approximately eight days. According to the company, the resulting system booted up successfully on a laptop, complete with a login screen, file storage, and support for networking and other basic functions.
In the fields of science and engineering, Anthropic claimed its models cracked two specific questions about the physics of black holes. First, the model found that black holes form just as abruptly in small universes as they do in large ones. Second, it allegedly found the limit of a technique that physicists use to rule out theories of quantum gravity in a simplified universe. Anthropic published research papers detailing both findings, though the results have not yet undergone independent peer review. Furthermore, a team of roughly 1,000 Fable 5.2 agents translated an existing mathematical proof of Fermat’s Last Theorem into a computer-verifiable form in just 30 hours, vastly outpacing an earlier version of Claude that took 11 days.
Access tiers and safety distinctions
While Fable 5.2 is available to the general public and paid subscribers across Anthropic platforms, Mythos 5.2 remains strictly controlled. Access to Mythos 5.2 is restricted to approved organizations through the company’s Cyber Verification Program and Life Sciences Verification Program. Anthropic explained that both products utilize the same underlying model architecture, but Fable includes stronger built-in restrictions concerning cybersecurity and biological research capabilities.
Comments