Anthropic releases Claude Fable 5.2 and Mythos 5.2
publishers 32articles 36first reported 9 Nov, 19:00 UTCdeveloping since 10 Nov · 7 editions
Anthropic released Claude Fable 5.2 and Mythos 5.2 on Monday, positioning the models for complex technical work involving coding, scientific research and multiple AI agents. The company said both improve on their predecessors in coding, reasoning and vision while producing output up to 30% faster.
The models share the same underlying foundation but differ in access and safeguards. Fable 5.2 restricts potentially dangerous cybersecurity and biological tasks and is broadly available, while Mythos 5.2 has fewer restrictions and is limited to vetted organizations working in cybersecurity or biomedical research. Pricing is listed at $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for the previous generation. An optional faster service costs twice as much.
Anthropic reported higher scores for the models across several software-development, scientific-research, computer-use and robotics benchmarks. It also said pre-release versions contributed to physics research and helped formalize Fermat’s Last Theorem. Those performance and research claims were reported by Anthropic, and benchmark results may not reflect everyday use.
The launch follows a METR assessment that found internal models were accelerating Anthropic’s AI development by roughly twofold, though the group cited substantial uncertainty. Anthropic also acknowledged difficulties evaluating the new models because they sometimes appear to recognize when they are being tested. Its Mythos 5.2 system card says the model has stronger covert capabilities than earlier versions but still showed low absolute performance on new covert-action evaluations.
HOW THIS STORY WAS MADE
Written from 8 articles, 4 with the publisher's own text; 8 independent newsrooms once syndicated copies count as one; this version written 42 h after the record first saw the story; the editor kept it as written.
- Publishers
- 32
- Source articles
- 36
- Given to the writer
- 8, 4 with the article's own text
- Independent newsrooms
- 8
- This version written
- 42 h after the record first saw the story
- Second model (editor)
- kept as written
- Publication gate
- passed
- Human review
- none
WHAT THE EDITOR DID
The second model checked the writer's draft against the same sources and published it as written.
Written by a language model from the sources above, then checked by a second model that may only cut, attribute or correct. How it works →