Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, upgrading its flagship model family designed for complex engineering challenges. During prerelease trials, a coordinated swarm of 1000 subagents worked continuously for 8 days to write, debug, and compile a functional laptop operating system completely from scratch. The updated software is available today on the Claude platform and major cloud providers, starting at $8 per million input tokens.
Anthropic assigned Fable 5.2 the task of building an operating system starting from an empty folder without human code generation. The model directed over 1000 subagents around the clock to write code, conduct tests, and review file structures. The resulting operating system boots on a laptop, complete with a custom kernel, disk and network drivers, file storage, a window management interface, a login screen, an installer, and integrated Claude functionality.
Fable 5.2 also demonstrated hardware engineering capabilities. Working from a single prompt, the model designed 3 working electronic devices, including a voice assistant hub and a USB speakerphone. It routed 4 layer circuit boards in under 24 hours, drafted computer aided design files for the enclosures, and wrote the device firmware without requiring electrical revisions. In pure mathematics, a team of 1000 Fable 5.2 agents formalized the proof of Fermat's Last Theorem in 30 hours, completing the task 7 times faster than previous attempts.
On the FrontierCode 2.0 agentic coding evaluation, Fable 5.2 scored 49.1%, outperforming Opus 5.5, GPT 6 Astra, and Gemini 4 Carbon. Output generation runs up to 30% faster than version 5.1. Standard access costs $8 per million input tokens and $40 per million output tokens. Anthropic also added a fast mode that accelerates generation speeds by 2.5 times, priced at $16 per million input tokens and $80 per million output tokens.
Anthropic separates the release into 2 deployment tiers sharing the same core model. Fable 5.2 features strict biology and cybersecurity safety classifiers for public access on Amazon Web Services, Google Cloud, and Microsoft Azure. Mythos 5.2 removes these limitations for verified research organizations participating in Anthropic cybersecurity and life sciences verification programs. Joe Haddad, a member of the technical staff at Vercel, described his experience testing the updated model.
The model is in my head. It understands the intent behind every prompt I have given it. I have not had to say No, I meant a single time. Projects we have been pushing along with other models for weeks got finished and approved by our teammates in a few days with Claude Fable 5.2, and the humans reviewing its code had less to say.
