Claude Fable 5.2 Is 20% Cheaper and Up to 30% Faster, Anthropic Says

Anthropic’s most powerful public AI model just got cheaper and quicker. Claude Fable 5.2, released on 9 November, costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1. It also writes answers up to 30% faster, the company says.

The same launch brought Claude Mythos 5.2. In its announcement, Anthropic says the two are “the same underlying model”. Fable carries extra safeguards that block most cybersecurity and biology tasks, which lets anyone on a paid plan use it. Mythos has looser limits in those areas and is open only to vetted organisations in Anthropic’s Cyber and Life Sciences Verification Programs, its documentation confirms.

What the numbers show

On FrontierCode 2.0, which checks whether a model’s code changes would be accepted by the people who run open-source projects, Fable 5.2 scored 49.1%. Claude Opus 5.5 scored 45.5%, OpenAI’s GPT-6 Astra 42.8% and Fable 5.1 32.4%, according to Anthropic’s own table. Google’s Gemini 4 Carbon still led one business-workflow test, 53.9% to 51.4%.

Anthropic also described its prerelease trials. About 1,000 Fable 5.2 agents formalised the proof of Fermat’s Last Theorem, rewriting it so a computer can check every step, in 30 hours. Its first attempt, with an unreleased research model, took 11 days. That follows earlier work in which AI agents checked maths’ longest proof. The company says the model also solved two open problems in black-hole physics and designed three working electronic devices.

For Mythos 5.2, Anthropic says an internal test found 10.2% more known software flaws than Opus 5.5 and 14.4% more than Mythos 5.1.

Who can use it

Fable 5.2 is available on the Pro, Max, Ultra, Team and Enterprise plans. Developers can reach it through the Claude API and on Amazon Bedrock, Google Cloud and Microsoft Foundry. It can take in up to 1 million tokens at once. An optional fast mode runs up to 2.5 times quicker at double the price.

Anthropic still tells most people to start with Opus 5.5. Dianne Penn, its head of product management for research, told CNBC that Fable 5.2 is meant for “power users or power use cases”. It is Anthropic’s first frontier release since CEO Dario Amodei called for a slower pace of AI development in September. “We called for pacing, not pausing,” Penn said. CNBC reports the company’s IPO could come as soon as next week.

On safety, Anthropic says Fable 5.2 wrongly blocked about 55% fewer requests than Fable 5.1 in early testing. It also has the company’s strongest protection yet against distillation, where rivals train models on another model’s answers, a problem that OpenAI has reportedly faced too.

Why it matters: A 20% price cut on Anthropic’s top public model makes long agent jobs, which can run for hours or days, cheaper to try. Every benchmark here comes from Anthropic, so independent tests will show how much of the lead holds.