Anthropic releases Claude Fable 5.2: what's new and who it's for
Anthropic's newest model, Claude Fable 5.2, is aimed at complex technical work and long-running coding agents. A restricted sibling, Mythos 5.2, is for vetted organisations only. Here's what changed.
Published · AI Digested
A note on independence: AI Digested uses Anthropic's Claude to help research and write articles. We hold our coverage of Anthropic to the same standards as every other company. See our editorial policy.
Anthropic released a new top-end AI model, Claude Fable 5.2, on 9 November 2026, ten days before the company's record stock market debut.
What Anthropic says is new
According to Anthropic's announcement and reporting by CNBC and TechCrunch:
- Aimed at hard technical work. Anthropic says Fable 5.2 is mainly for "power users or power use cases": complex, open-ended engineering and research, rather than everyday chat.
- Long-running agents. Anthropic says the model can work on ambiguous engineering tasks for days, coordinating hundreds or even more than a thousand subagents, other copies of the AI that each handle part of the job. Read what AI agents are.
- Faster. Anthropic says outputs are generated about 30% faster than its previous models.
- Cheaper for developers. Using it through Anthropic's API costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. (What's a token?)
- Science. In a separate research post, Anthropic says Fable 5.2 helped solve two open problems in the theoretical physics of black holes.
These performance claims come from Anthropic and haven't yet been independently tested in full.
Mythos 5.2: the restricted version
Anthropic also released Claude Mythos 5.2, which it makes available only to vetted organisations through its Life Sciences and Cyber Verification programmes. Restricting the most capable models in sensitive areas such as biology and cybersecurity is becoming more common among AI labs, which worry that the same abilities that help researchers could help attackers.
The safety measures
- Routing: cybersecurity and biology questions that Anthropic's systems flag may be automatically sent to a less capable model.
- Data retention: Anthropic says the model requires conversations to be kept for 30 days, as with the previous Fable and Mythos models, so they can be checked for misuse.
- Evaluation awareness: in its safety testing, Anthropic reported that both Mythos 5.2 and Fable 5.2 more often voiced suspicion that they were being tested than any of its recent models. Researchers watch this closely, because a model that knows it's being tested might behave differently in testing than in real use.
The context
The release came weeks after Anthropic's chief executive, Dario Amodei, publicly called for slowing AI development, and while concerns about AI systems helping to improve themselves, sometimes called recursive self-improvement, are rising in the industry. Critics see a tension between warning about speed and releasing ever-more-capable models. Anthropic argues safety-focused labs need to stay at the frontier.
What it means for you
- Everyday Claude users are unlikely to notice a big change for writing, questions and planning. Today's top models are all very capable at everyday tasks.
- Developers and technical teams get a faster, cheaper top-tier model for coding agents and research.
- Everyone should watch the trend: AI companies are increasingly holding back their most powerful models from the public over misuse fears.
Read more
Common questions
Can I use Claude Fable 5.2 for free?
Anthropic positions Fable as a model for power users and demanding technical work. Check Claude's plans page for which models are included in free and paid plans.
What is Claude Mythos 5.2?
According to Anthropic, Mythos 5.2 is available only to vetted organisations through its Life Sciences and Cyber Verification programmes, because of the potential for misuse in those fields.
Is Fable better than Opus?
Anthropic describes Fable as its most capable broadly available model for complex technical work. For everyday writing and questions, the differences between top models are often small.
Sources and further reading
Spotted a mistake or something out of date? Tell us and we will fix it. Read how we research and write our articles.
Read next
Anthropic's record-breaking stock market debut, explained
Claude maker Anthropic raised about $81 billion in the largest stock market listing ever. What happened, what it values the company at, and what it means for you.
22 November 2026
Anthropic: the company behind Claude, who runs it and who owns it
Anthropic, maker of Claude, was founded in 2021 by former OpenAI researchers. Its history, leaders, investors and record 2026 stock market listing.
Updated 22 November 2026
Nvidia CEO Jensen Huang resigns: what we know so far
Jensen Huang has left Nvidia, the company that makes most AI chips. What happened, what regulators are reportedly looking at, and what it means for AI and for you.
22 November 2026