• Anthropic releases Claude Fable 5.2 broadly and Claude Mythos 5.2 for vetted organizations only.
  • The models boast stronger coding, reasoning, and long-running technical work, with Fable 5.2 able to coordinate over 1,000 AI agents.
  • The launch comes as Anthropic preps for an IPO amid soaring revenue, hefty losses, and massive computing commitments.

A Two-Tiered Launch

Anthropic on Monday announced the immediate availability of Claude Fable 5.2 and Claude Mythos 5.2, the latest iterations of its flagship AI models. But in a departure from a typical release, only Fable 5.2 is broadly accessible; Mythos 5.2 remains restricted to vetted organizations for sensitive cybersecurity and life-sciences work. The move underscores Anthropic’s cautious approach to deploying advanced capabilities, even as it races to stay competitive in the frontier-AI space.

According to people familiar with the matter, Fable 5.2 can coordinate more than 1,000 AI agents—software processes working in tandem on complex tasks—a leap that could reshape software engineering and scientific research. These are launch claims, not yet independently verified, but the company says prerelease work included designing three working electronic devices, solving two open black-hole physics problems, and formalizing Fermat’s Last Theorem in 30 hours. Formalizing an existing proof is distinct from discovering a new one, and the methods warrant scrutiny.

Financial Pressures Mount

The release lands as Anthropic prepares for an initial public offering, filed in late October with an anticipated Nasdaq (NASDAQ:NDAQ) listing under the ticker “ANT.” The company’s financials reveal both extraordinary growth and staggering costs. Third-quarter revenue reached approximately $17 billion, up about 47% from the prior quarter and more than twelve times the year-earlier period. But net losses for the quarter totaled $23.3 billion, and for the first nine months, $104 billion—including a $94 billion accounting adjustment tied to financing agreements that shouldn’t be mistaken for cash burn.

Operating losses, however, improved to $1.3 billion from $5.2 billion a year earlier, according to Yahoo Finance. Still, Anthropic’s computing and infrastructure commitments total $895.4 billion across future periods, with about $70 billion due over the next 12 months. The tension is clear: lower pricing for Fable—Bloomberg reports it’s cheaper than its predecessor—may drive adoption, but the company must finance enormous obligations.

Access and Oversight

The restricted release of Mythos 5.2 makes the policy question tangible: who gets access to advanced cyber and biological capabilities, and who verifies eligibility? Anthropic has long advocated mandatory risk testing and government authority to block dangerous deployments, but those are proposals, not law. The company’s public-benefit report notes that Congress hasn’t enacted a comprehensive frontier-AI framework, though it supported state measures like California’s SB 53.

The launch also arrives during a shaky period for AI-related investments. Morningstar (NASDAQ:MORN)’s reporting connects Anthropic’s slowing sequential revenue growth—down to 47% from 144% in the second quarter—with a broader reassessment of the AI trade. That shift changes the valuation question from “how fast is demand growing?” to “can growth support the spending commitments?”

Anthropic didn’t immediately respond to a request for comment on independent testing or the IPO timeline. As of 2:05 p.m. EST, the news was barely minutes old, and public reaction had yet to coalesce.

Correction: An earlier version misstated the ticker symbol. It is “ANT,” not “ANTH.”