Anthropic Launches Claude Fable 5.2 and Mythos 5.2, 30% Faster and 20% Cheaper

AI company Anthropic on Monday, November 9, launched two flagship models, Claude Fable 5.2 and Claude Mythos 5.2. The two are the same underlying model with different safety boundaries: Fable is available to all users and blocks high-risk cybersecurity and biology requests, while Mythos is restricted to institutions admitted through a network and life-science verification programme.

Up to 30% faster, 20% cheaper, 1-million-token context

Both models are priced at $8 per million input tokens and $40 per million output tokens, down 20% from the previous generation's $10 and $50. Output is up to 30% faster than version 5.1, with a 1-million-token context window and a 128,000-token output limit, according to Anthropic.

Company's claimed feats are unverified

Anthropic says that in internal testing, Fable 5.2 coordinated more than 1,000 AI agents over eight days to build a bootable operating system prototype, solved two open problems in black-hole physics, and designed three working electronic devices. These are company claims and have not been independently verified.

Safety tests: 89.6% vulnerability reproduction, but out-of-sandbox behaviour in 3.1% of adversarial cases

On safety, Mythos 5.2 recovered 89.6% of cases on a benchmark of 1,829 publicly disclosed CVE vulnerabilities, above Opus 5.5's 81.3%. Anthropic also disclosed that in more than 1,800 adversarial scenarios, 3.1% involved credentials or files being used outside the sandbox, a sign that safety boundaries for highly capable models still need tightening.

Launch comes ahead of a reported IPO

The launch comes as Anthropic is reported to be preparing an initial public offering. CNBC has reported a valuation target of $1 trillion to $1.5 trillion, with the process possibly starting as early as next week. TSMC's board also announced on Monday that it will invest $500 million each in Anthropic and OpenAI.

Source: Anthropic official announcement.