Anthropic releases Fable 5.2 and Mythos 5.2 with API prices cut by 20 per cent

Anthropic has released Claude Fable 5.2 and Claude Mythos 5.2, lowering standard API token prices by 20 per cent and claiming output up to 30 per cent faster than the 5.1 generation. The company announced the models on Monday, and its documentation dates the release 9 November.

The two models share one underlying model and differ in safeguards and who may use them. Fable 5.2 blocks most cybersecurity and biology tasks, which allows Anthropic to offer it to all users. Mythos 5.2 has fewer of those restrictions and is open only to vetted organisations in cybersecurity or biomedical research. Fable 5.2 can be reached through the Claude API and through Amazon Bedrock, Google Cloud and Microsoft Foundry, though arrangements differ by platform.

Both models list at $8 per million input tokens and $40 per million output tokens, compared with $10 and $50 for their predecessors. A fast mode that Anthropic says runs up to 2.5 times quicker is priced at $16 and $80 per million input and output tokens respectively. The speed figures are company claims, and the final bill depends on how much reasoning, how many agents and which tools a job uses. Fable 5.2 handles a context window of one million tokens and up to 128,000 output tokens.

On Anthropic's own FrontierCode 2.0 coding test, Fable 5.2 scored 49.1 per cent at Xhigh effort, against 32.4 per cent for Fable 5.1 at the same setting. At High effort it reached 45.3 per cent, near the 45.5 per cent recorded by Opus 5.5 at Max, at about half the cost per task. Fable 5.2 scored lower at its Max setting than at Xhigh, and all results are reported by the vendor.

In a separate research report, Anthropic said it had Fable 5.2 design a USB speakerphone, a home voice-assistant hub and a handheld virtual pet with manufacturer Diode Computers. Anthropic says the prototypes functioned with no electrical changes but have not been through electrical validation, antenna tuning, thermal testing or certification. Professor Ethan Mollick, who tested the model early, described near-expert performance on many computer-based tasks, but said one creative project stalled after agent teams produced an 87,500-word rulebook. "The AI had built a bureaucracy," he wrote.

Developers upgrading from 5.1 face changes, as Anthropic's documentation warns that editing earlier conversation history can invalidate retained reasoning blocks. An older computer-use interface no longer works on most platforms, though Amazon Bedrock is an exception. Data is kept for 30 days unless Anthropic expressly approves zero data retention.

The first Fable and Mythos models were withdrawn within days of their June launch after the US Commerce Department ordered Anthropic, on national security grounds, to bar foreign nationals. Those controls were lifted weeks afterwards. Anthropic is targeting an IPO in the week of 16 November, when it is expected to raise up to $100 billion, according to Yahoo Finance. A securities filing showed third-quarter revenue of $17 billion and an operating loss of $567 million.