Anthropic releases Claude Fable 5.2 and Mythos 5.2, cuts prices
Anthropic’s first frontier models since Dario Amodei’s September call to slow AI development cost 20% less than the versions they replace.
Summary
In brief
- Claude Fable 5.2 costs $8 per million input tokens, 20% less than the 5.1 models
- Anthropic targets a $1 trillion to $1.5 trillion valuation in its IPO, CNBC reported
- Claude Mythos 5.2 is limited to vetted cyber defense and biomedical research groups
Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on Monday, new versions of its most advanced models that the company said generate output up to 30% faster and cost 20% less than their predecessors.
It is Anthropic’s first release of frontier models, the most capable systems a lab offers, since Chief Executive Dario Amodei called in September for a slower pace of AI development, according to CNBC’s report. The company is also preparing an initial public offering at a targeted valuation of $1 trillion to $1.5 trillion, CNBC reported.
“We called for pacing, not pausing,” Dianne Penn, Anthropic’s head of product management for research, told CNBC. Anthropic said it developed new evaluations for behaviors such as sandbox escapes, in which a model breaks out of its restricted test environment, and that the UK’s AI Security Institute and METR tested the model before release.
Fable and Mythos are the same underlying model. Fable has cybersecurity and biology safeguards and is available to all users, while organizations working on cyber defense or biomedical research can apply for Mythos through Anthropic’s Life Sciences and Cyber Verification Programs. Both are priced at $8 per million input tokens and $40 per million output tokens, the units of text a model reads and writes.
Anthropic said in its announcement on X that the models are a step above Claude Opus 5.5, its everyday model, in almost all the capabilities it evaluated. On FrontierCode 2.0, a test of multistep coding work done autonomously, Fable 5.2 scored 49.1% at its Xhigh effort setting, against 45.5% for Opus 5.5 and 42.8% for OpenAI’s GPT-6 Astra. The results have not been independently verified.
The company also said 1,000 Fable 5.2 agents formalized Fermat’s Last Theorem in 30 hours, translating the proof into a language a computer can check step by step, while Opus 5.5 completed 26% in 18 hours. In a post on X, Anthropic said Mythos 5.2 accurately modeled 28 of the 34 capsids it attempted, the protein shells used to deliver gene therapies.
Anthropic introduced Mythos in April, limiting access to a select group of organizations because of its advanced cybersecurity capabilities, and released Fable, a public version with added safeguards, two months later. The company said in its amended prospectus that revenue for the nine months to September 30 was $33.24 billion, up about 1,200% from a year earlier.
Executives are preparing to start the IPO roadshow, and the offering could take place as soon as the week of November 16, CNBC reported.