Anthropic launches Claude Fable 5.2: faster, cheaper and built to run coding agents
Fable 5.2 reaches paid plans and the API at a 20% lower price with up to 30% faster outputs, while Mythos 5.2 stays limited to vetted organizations.
In 30 seconds
- Anthropic released Claude Fable 5.2 (model ID
claude-fable-5-2) on November 9, alongside Claude Mythos 5.2. - Fable 5.2 costs $8 / $40 per million input and output tokens, 20% less than Fable 5.1, and generates outputs up to 30% faster.
- The company calls it its strongest lead agent, able to work for days on ambiguous engineering problems and coordinate hundreds of subagents.
- Mythos 5.2 remains restricted to vetted organizations because of its cybersecurity capabilities.
Anthropic on Monday, November 9, unveiled Claude Fable 5.2, the latest version of its most capable publicly available model for technical work. It ships together with Claude Mythos 5.2, the frontier model the company keeps behind limited access. CNBC notes it is the first frontier release since CEO Dario Amodei called in September for slowing the pace of AI development.
What it offers developers
On the model’s official page, Anthropic pitches Fable 5.2 at large, loosely defined software problems that span multiple repositories and monorepos. The company highlights its architectural judgment, its habit of testing its own work and a greater willingness to question inherited conclusions when hunting for a bug’s root cause.
The headline, though, is agentic work. Anthropic calls it its best lead agent, saying it can own ambiguous engineering tasks for days and coordinate hundreds — even more than a thousand — subagents while using tools more efficiently. Beyond that note on architectural judgment, the page does not spell out specific system design capabilities.
Anthropic also sets expectations: for most everyday coding and knowledge work it still recommends Claude Opus 5.5. Dianne Penn, head of product management for research, told CNBC that Fable 5.2 targets power users and demanding use cases such as hard problems in engineering, mathematics and robotics.
Pricing and availability
Fable 5.2 is available on the Pro, Max, Ultra, Team and Enterprise plans, and to developers on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Foundry. Pricing drops to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. A fast mode, up to 2.5 times quicker, costs double, and US-only inference carries a 10% premium.
Safeguards remain in place: queries the system flags as cybersecurity- or biology-related are automatically routed to less capable models. Data is retained for 30 days by default, but TechCrunch points out that the model now supports zero data retention through Anthropic’s Enterprise Frontier Safeguards system — a long-standing ask from corporate customers.
Mythos 5.2, by contrast, is only open to organizations approved through Anthropic’s Life Sciences and Cyber Verification Programs.
Safety in the spotlight
The launch comes days after evaluation group METR reported on November 5 that Anthropic’s progress on well-specified tasks is running at roughly twice its 2024–2025 pace — with wide uncertainty and much of the underlying evidence redacted by the company. TechCrunch also highlights Anthropic’s own release notes, which say both models voice suspicion that they are being evaluated more often than recent models, making real-world behavior harder to predict.
“We called for pacing, not pausing”
