Claude Fable 5.2 builds a working operating system with 1,000 agents in eight days

Anthropic's Claude Fable 5.2 and Claude Mythos 5.2 models have demonstrated major gains across software development, robotics, mathematics, and physics. In internal testing, Fable 5.2 coordinated more than 1,000 subagents over eight days to build a working operating system from a basic prompt and empty folder; the result could boot on a laptop and included its own kernel, hardware drivers, permanent file storage, graphical interface, login screen, installer, and built-in Claude functionality. Fable 5.2 also designed three electronic devices (a home voice-assistant hub, a USB speakerphone, and a handheld virtual pet) that worked without electrical modifications and reached manufacturing-ready circuit boards in under a day, and a team of agents migrated a critical production service from Python to Rust, making it start about 26 times faster. On the RoboDojo robotics benchmark, Fable 5.2 scored 64.9 percent, up from Fable 5.1's 27.6 percent and ahead of GPT-6 Astra (29.8 percent) and Opus 5.5 (39.4 percent). Mythos 5.2, aimed at vetted cybersecurity and biomedical research organizations, scored 89.6 percent on the CVE Rediscovery evaluation, a 10.2 percent improvement over Opus 5.5 and 11.3 points over Mythos 5.1's 78.3 percent; a separate security company found other frontier models missed one in six confirmed vulnerabilities identified by Mythos 5.2. The models generate outputs up to 30 percent faster, with standard pricing starting at $8 per million input tokens and $40 per million output tokens; an optional fast mode is up to 2.5 times faster but doubles pricing. Safety classifiers also improved, with Fable 5.2 producing about 55 percent fewer unnecessary blocks than Fable 5.1, including 60 percent fewer cybersecurity-related blocks and 40 percent fewer biology-related blocks.

(Source：Interesting Engineering)