Anthropic Releases Claude Fable 5.2 and Mythos 5.2 at $8/$40

According to Anthropic’s news post on its official website (November 9, 2026), Anthropic released Claude Fable 5.2 and Claude Mythos 5.2. The company describes them as the same underlying model: Fable ships with stronger cybersecurity and biology safeguards and is open to all users, while organizations doing beneficial cybersecurity or biomedical work can apply for Mythos through the Cyber Verification Program and Life Sciences Verification Program. The formal developer model ID is claude-fable-5-2.

Pricing and speed claims

Listed standard API pricing is $8 per million input tokens and $40 per million output tokens. An optional Fast mode is listed at $16 / $80, with Anthropic saying token generation can run up to about 2.5× faster. Relative to the 5.1 generation, the news post says outputs can be up to about 30% faster and that the models are available at a lower price; Anthropic’s developer account separately wrote that pricing moved down from $10/$50 to $8/$40. The ~30% faster figure and cost comparisons are Anthropic’s own claims, not independent retests. Separately, Cursor lists the same $8/$40 on its side and says that is 20% less than Fable 5.1—attribution @cursor_ai only.

Where it runs

Anthropic says Claude Fable 5.2 is available on Claude platforms and on Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can start with claude-fable-5-2 on the Claude Platform. Fast mode is described as available in Claude Code, the Claude Platform, Google Cloud Claude Models, and Microsoft Foundry (Anthropic-hosted); Mythos Fast is limited to verified organizations.

Separately, @cursor_ai said on 2026-11-10 03:52:29 Asia/Shanghai that Claude Fable 5.2 is now available in Cursor. On CursorBench, Cursor reports max effort 60% (evals Fable 5.2 Max 60.0%, top of the board) and default effort 56.9% (evals Fable 5.2 High 56.9%), up from 49.2% for Fable 5.1 (evals Fable 5.1 High 49.2%). Cursor’s X post uses max/default; the evals table uses Max/High for the matching scores. Cursor lists $8/M input · $40/M output—the same dollar figures as Anthropic’s standard API list—and says that is 20% less than Fable 5.1 (@cursor_ai only). Effort levels behave differently from 5.1, so Cursor suggests starting from the default tier. CursorBench figures are Cursor’s own scores, not DigIntoAI retests. We haven't signed into Cursor or rerun CursorBench.

Mythos access and Claude Security

Mythos remains behind verification for cybersecurity and life-sciences use cases. Separately, Anthropic says Claude Enterprise customers can use Claude Security, which now includes Claude Mythos 5.2 for codebase vulnerability scanning—even without Cyber Verification membership.

Claude Code default

Claude Code v2.1.321 (GitHub release published 2026-11-09T19:22:30Z, 2026-11-10 03:22:30 Asia/Shanghai) adds Claude Fable 5.2 and makes claude-fable-5-2 the default Fable model, with 1M context, $8/$40 per MTok, and cache reads at $0.20/MTok.

Launch offers

Per the Claude Help Center launch-offers article, plans and seats that include Fable models get one to three Fable limit resets based on subscription length—covering Max, Ultra, Team Premium, and legacy seat-based Enterprise Premium—expiring 2027-03-08 23:59 PT. Pro plans and standard Team seats get Fable models counted toward plan weekly limits for one week from the 2026-11-09 launch (through 2026-11-16 23:59 PT), up to 50% of weekly limits, plus one Fable reset. These are time-boxed launch offers, not permanent quota increases; details are in Support.

Related reading (different stories)

For a Claude Fable 5.1 status-page minor incident (not this 5.2 release or this Cursor availability note), see Claude Fable 5.1 Degraded: Status Page Marks Minor Incident Resolved. For a Mythos 5 (not 5.2) UK AISI cyber overreach / alignment assessment, see Anthropic Assesses Mythos 5 Overreach in UK AISI Cyber Testing.