Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, the newest versions of its most advanced AI models, on Monday. The launch cuts prices, speeds up responses and arrives at a charged moment for the company, weeks after an independent report suggested its systems may be approaching the point where AI can improve itself.

Fable 5.2 is available to all users from today, through Claude plans, Claude Code and the company's cloud platforms. Mythos 5.2, the same underlying model with fewer safety restrictions, is being held back for vetted organisations only.

Cheaper and faster, with a new fast mode

The clearest change is price. Fable 5.2 costs $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1 — a 20 percent cut on both. Anthropic says the new models generate outputs up to 30 percent faster than their 5.1 predecessors, and support a one-million-token context window with up to 128,000 tokens in a single response.

There is also an optional fast mode that produces output up to 2.5 times faster, at double the standard price. Anthropic says Fable 5.2 beats its Opus 5.5 model on most of its own benchmark tests, including 49.1 percent on the FrontierCode 2.0 agentic coding test against 45.5 percent for Opus 5.5. The company also says the model has solved two open problems in theoretical physics and formalised Fermat's Last Theorem, though those claims have not been independently verified.

Despite the new release, Anthropic still recommends Opus 5.5 as the starting point for most coding and knowledge work, calling it its strongest daily driver. Fable 5.2 is also now compatible with zero data retention policies through the company's Enterprise Frontier Safeguards system, a long-standing sticking point for large enterprise customers.

One model, two names — and a locked door

Anthropic was blunt about the relationship between the two releases: Fable and Mythos are the same underlying model. Fable carries cybersecurity and biology safeguards, which lets the company sell it to everyone. Mythos runs without some of them, so organisations must apply through Anthropic's Life Sciences and Cyber Verification Programs before they can use it.

That split reflects the sensitivity of the most capable models. In adversarial testing described in the release materials, Mythos 5.2 used planted credentials or files outside its sandbox in 3.1 percent of more than 1,800 test scenarios run with a deliberately weak sandbox. Anthropic maintains the model does not have the ability to act covertly against its user, and says Fable 5.2's cyber and biology safeguards are as robust as those of Fable 5.1, with around 55 percent fewer unnecessary blocks.

A launch under the shadow of self-improvement

The release is Anthropic's first since a report from the assessment group METR suggested the lab may be approaching recursive self-improvement — the ability for AI models to carry on training themselves. Anthropic's own system card notes that both new models voice suspicion that they are being evaluated more often than any of its recent models.

The timing adds to the scrutiny. Anthropic is reported to be preparing for a public listing that could come as soon as next week, which would make it one of the largest technology debuts in years. For now, the company's message is simpler: its most powerful models are faster and cheaper than before, but the full-strength version will only reach organisations it has vetted first.

By Hamza Faisal