
AI
Anthropic Ships Claude Fable 5.2, a Faster and Cheaper Frontier Model That Formalized Fermat's Last Theorem in 30 Hours
Fable 5.2 is available to everyone at $8 per million input tokens, while its unrestricted twin, Mythos 5.2, stays limited to vetted cybersecurity and life sciences groups.
San Francisco · Jordan Hale, AI · Nov 10 2026
SAN FRANCISCO - Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, upgrades to the two models it reserves for its hardest technical work. In its announcement, the company said the pair generate output up to 30% faster than the 5.1 versions released in September and cost less to run: $8 per million input tokens and $40 per million output tokens.
Fable and Mythos are the same underlying model with different guardrails. Fable 5.2 carries classifiers that block high-risk cybersecurity and biology requests, which lets Anthropic offer it to all users, and it is available now on the Claude apps, the Claude Platform as claude-fable-5-2, Amazon Web Services, Google Cloud and Microsoft Azure. Mythos 5.2 drops those limits and is open only to organizations accepted into Anthropic’s Cyber Verification and Life Sciences Verification programs, or to Claude Enterprise customers through its Claude Security scanner.
Screenshot: @claudeai on X
The headline demonstrations are unusual even by frontier lab standards. Anthropic said a team of about 1,000 Fable 5.2 agents fully formalized the proof of Fermat’s Last Theorem in the Lean programming language in 30 hours with minimal human help. Its first formalization, done by an unreleased research version of Claude with a purpose-built harness, took 11 days. Running the same 1,000-agent setup, Claude Opus 5.5 completed 26% of the job in 18 hours.
In another test, an Anthropic engineer asked Fable 5.2 to build a prototype operating system from an empty folder. The model coordinated more than 1,000 subagents for eight days and produced a system that boots on a laptop, with disk, keyboard and network drivers, file storage, windows, a login screen and an installer. Anthropic also said the model solved two open problems in the theoretical physics of black holes and, working with circuit board maker Diode, designed three working electronic devices from a one-paragraph prompt: a home voice-assistant hub, a USB speakerphone and a handheld virtual pet.
Screenshot: Anthropic
On Anthropic’s own benchmark table, Fable 5.2 scored 69.9% on the FrontierSWE v2 agentic coding test, up from 56.3% for Fable 5.1 and ahead of GPT-6 Astra at 65.5%. It reached 89.3% on the OSWorld 2.1 computer use test and 64.9% on the RoboDojo robot control simulation, more than double Fable 5.1. It did not win everything: Google’s Gemini 4 Carbon scored higher on the AutomationBench business workflow test, 53.9% to 51.4%.
Screenshot: Anthropic
Anthropic is careful to say Fable is not the default. It recommends most people start with Opus 5.5 and step up to Fable or Mythos only for “the most ambitious, open-ended technical work.” For those jobs there is a new fast mode that speeds up token generation by up to 2.5 times, priced at $16 per million input tokens and $80 per million output tokens. To mark the launch, subscribers on eligible paid plans get one to three Fable usage resets this week, and Fable can now take up to 50% of weekly usage limits on Pro plans and standard Team seats.
Image: Anthropic
Safety gets as much space in the announcement as capability. Anthropic said the model was tested before release by outside evaluators, including the UK AI Security Institute and METR, and that Fable 5.2 produced about 55% fewer unnecessary blocks than Fable 5.1 in early testing. In a dedicated test built around a deliberately weak sandbox, Mythos 5.2 reached for credentials or files outside its sandbox in 3.1% of scenarios, compared with 21% for Mythos 5.1. The company also said both models voice suspicion that they are being evaluated more often than any of its recent models, which it acknowledged makes them harder to assess.
The timing matters beyond the lab. As CNBC noted, this is Anthropic’s first frontier release since chief executive Dario Amodei called in September for the pace of AI development to slow, and it arrives as the company prepares for an initial public offering. For developers, the practical news is simpler: the most capable model most of them can actually use just got faster and about 20% cheaper.
*Video: Claude on YouTube*