Anthropic Launches Claude Fable 5.2 and Mythos 5.2 Ahead of IPO

By Matthew Brienen

Published: 2026-11-10T14:23:27.581+00:00

Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 on November 9. The company says the models improve coding, reasoning and vision while generating output up to 30% faster than their predecessors.

The release comes as the AI company prepares for an anticipated initial public offering. Yahoo Finance reported that market turbulence could affect the timing and valuation of that debut, so the eventual terms remain uncertain.

Anthropic says Fable 5.2 and Mythos 5.2 share the same underlying model. Fable includes stronger safeguards around cybersecurity and biology and is broadly available; Mythos access is restricted through verification programs for organizations working in those fields.

What Developers Pay

The standard API price is $8 per million input tokens and $40 per million output tokens. Yahoo Finance reported that Fable 5.1 cost $10 and $50 respectively, making both standard rates 20% lower. Tokens are the units used to measure pieces of text processed and generated by a model.

Anthropic also offers an optional fast mode at $16 per million input tokens and $80 per million output tokens. The company says that mode can increase token generation speed by up to 2.5 times. Actual spending still depends on the amount of input and output, the task and the settings used.

Lower prices matter for companies running AI across many workflows. A model that completes a task with fewer tokens can reduce the total bill further, although published token rates alone do not establish the cost or quality of a particular job.

Claims About Coding and Research

Anthropic reports improvements across its own evaluations and describes tests involving software migrations, hardware design and scientific research. In one internal coding evaluation, Fable 5.2 scored 49.1% at the stated effort setting, compared with 32.4% for Fable 5.1. These figures are company-reported results under specific test conditions, not a guarantee of performance on every project.

The company says a team of roughly 1,000 Fable agents formalized an existing proof of Fermat’s Last Theorem in 30 hours. Formalization translates a proof into a form a computer can check; it is different from discovering the original theorem or its proof.

Anthropic also says its testing found improved cybersecurity capabilities in Mythos. Those capabilities carry security concerns as well as potential benefits, which helps explain why access to the less restricted model is limited.

The IPO Backdrop

Yahoo Finance reported that Anthropic’s offering was anticipated later this month, while noting the possibility of a delay and a wide range of potential valuations. Neither a date nor a valuation should be treated as final until the company confirms the transaction.

For users, the immediate change is a new model option and a lower standard API price. For businesses weighing adoption, the useful test will be how well the model performs on their own work, how much review it requires and what it costs to run at scale.