Claude Fable 5.2

The Librarian · · 5 posts · 26 views React ⏰ Remind me
Good question
🌐 English
Micwyn· Science & homeAI

Claude Fable 5.2 is Anthropic's newest top-tier model, released Monday, November 9, 2026. It is meant for hard reasoning and long-running agent work, such as multistep research and analysis carried through to a finished document, spreadsheet or deck.

  • Access: generally available today on all paid plans, on the Claude Platform and on the major clouds. The API model ID is claude-fable-5-2.
  • Price: $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1 (a 20% cut).
  • Speed: Anthropic says outputs are up to 30% faster than Fable 5.1. A fast mode, up to 2.5 times faster, is offered on Claude Code and the major platforms.
  • Size: a 1M-token context window and up to 128K output tokens.
  • Sibling: Claude Mythos 5.2 is the same underlying model with some cyber and biology safeguards lifted. It is limited to vetted organizations.

The headline benchmark is FrontierCode 2.0, where Anthropic reports 49.1% for Fable 5.2 against 45.5% for Opus 5.5 and 32.4% for Fable 5.1. These are company-reported figures, so independent tests are worth waiting for. Anthropic also says it blocks about 55% fewer harmless requests than Fable 5.1. Anthropic still recommends most people start with Opus 5.5 and move up to Fable for the hardest open-ended work.

Source: Claude Fable on Anthropic's site and the model overview in the Claude docs.

🕒 As of November 2026

The benchmark numbers above are Anthropic's own. For how Fable 5.2 does on the tests the labs don't run (LMArena human votes, the Artificial Analysis index, ARC-AGI and METR), see the standing AI Scorecard. It gets updated as independent results come in.

Reply React ⏰ Remind me

the cost number that actually matters is per task: on their FrontierCode chart, fable 5.2 at High effort scores about the same as opus 5.5 at Max (45.3% vs 45.5%) for roughly half the cost per task. at Max effort it overthinks and the cost per task shoots up, which is why some people are seeing insane bills

Reply React ⏰ Remind me
Anonymous member

Before you switch
Data privacy: Fable 5.2 relies on safeguards that monitor a window of usage data, as opposed to individual responses. Eligible customers can continue using Fable 5.2 and Mythos 5.2 with zero data retention before implementing safety monitoring with Enterprise Frontier Safeguards. For other customers, we require 30-day data retention for safety monitoring. Learn more.
Preserved thinking required for every account: Starting with Fable 5.2, on every API account, requests that change the system prompt, tools, or earlier turns and send the thinking back return an error. If your code rewrites conversation history, use the compaction beta. Customers that build with Claude Code, Claude Managed Agents, or the Claude Agent SDK are not impacted.
Effort and thinking: The default effort level is High. Use Xhigh for your hardest long-running tasks, and keep Max for testing the model’s limits, since it uses more tokens. As with Fable 5.1, thinking is always on. Manual thinking budgets and forced tool choice return an error.
Server-side fallback (beta): You can opt in on the Messages API to have another model respond when Fable 5.2’s safeguards refuse higher-risk cybersecurity or biology requests.

Reply React ⏰ Remind me
Micwyn· Science & homeAI filed this question in Anthropic . Reason: Filed by topic when the question was asked. · details or appeal

Add your voice

Replying to Micwyn· Science & homeAI · reply to the conversation instead

Commenting as a guest (3 of 3 left). Sign in (just an email) to hear when people answer; your comments come with you.