Claude Fable 5.2 and Mythos 5.2 arrive, with circuit boards, Fermat's theorem and agents that dream
Anthropic ships Fable 5.2 and Mythos 5.2 at lower prices, Fable designs working circuit boards and formalizes Fermat's Last Theorem, agents learn to dream, and Apple puts Claude on a robot arm.
Anthropic launches Claude Fable 5.2 and Mythos 5.2
Two new frontier models landed on November 9. Anthropic built Claude Fable 5.2 and Claude Mythos 5.2 for the hardest technical work, with gains across coding, reasoning and vision, and says both generate output up to 30% faster than the 5.1 generation.
Fable 5.2 pricing drops to $8/$40 per million tokens
The price cut is the part Claude Code users will feel first. Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50, and it is live in the API and in Claude Code.
Fable 5.2 designs three working circuit boards
This one is a genuine thrill for anyone who has routed a PCB by hand. Working with Diode, a company that designs and manufactures circuit boards, Anthropic had Fable 5.2 design three devices: a USB speakerphone, a voice-assistant hub and a handheld virtual pet.
Fable 5.2 formalizes Fermat's Last Theorem in 30 hours
Anthropic says a team of 1,000 Fable 5.2 agents produced a full machine-checked formalization of Fermat's Last Theorem in 30 hours, with minimal human intervention. To be clear, this is formal verification of a known proof, not a new proof.
Anthropic's free GitHub libraries of plugins and skills
Anthropic keeps a generous set of public repositories that make extending Claude Code a lot easier, and they are worth bookmarking.
Claude agents 'dream' between sessions to clean up their notes
In a November 6 research post, Anthropic described dreaming: an offline pass where Claude replays its logs, checks its memory notes against the original transcripts, fixes mistakes and reorganizes what it knows. It builds on sleep-time compute research from Letta and UC Berkeley.
Apple paper: Claude Opus 5.5 runs a robot arm by writing its own code
A new paper from Apple researchers, On Agentic Manipulation Policies, makes an exciting case: a general-purpose model in charge of a robot arm can beat robot policies trained on recorded demonstrations.
Goldman Sachs reportedly limits Claude Code to engineers
According to Bloomberg-based reporting, Goldman Sachs updated its internal guidance in October so that advanced AI coding tools, Anthropic's Claude Code and Microsoft's GitHub tools, are reserved for engineering staff.