Effort defaults to high in the Claude Platform and medium in Claude apps. The xhigh setting is meant for hard, long-running agentic work with measurable gains. Max is for finding the model's ceiling and is not recommended for everyday use.
Fable 5.2 is now available in the Claude Platform and Claude Code. It is up to 30% faster than Fable 5.1 and priced at $8/$40 per 1M input/output tokens, down from $10/$50.
Fable 5.2 generates output up to 30% faster than 5.1 and is available on all paid plans and major cloud platforms. Addy Osmani notes Fable is best for ambitious open-ended technical work, while Opus 5.5 remains the recommended daily driver for most coding and knowledge tasks.
Fast mode is available for Fable 5.2, delivering 2.5x faster output at twice the cost. In the API, enable it with `"speed": "fast"`; in Claude Code, toggle it in the model picker, which bills through extra usage credits.
With Fable 5.2, Anthropic is rewarding subscriber loyalty: the longer someone has been subscribed, the more Fable resets they receive. Ultra and Team Premium users get up to 3 resets; Pro and Team Standard get 1 reset plus a trial week.
Using a Fable reset restores only the Fable portion of a user's weekly limit; usage on other models within that week remains unchanged.
Pro subscribers get one Fable reset; Max, Ultra, and Team Premium users get up to three, with the exact count scaling by how long they've been subscribed.
A demo shows Fable 5.2 generating a spider web with 7,330 individual dew drops entirely in code, with no image files. Each drop simulates accurate light refraction, showing the garden reflected upside down inside it.
A demo shows Fable 5.2 writing a compiler that lets you author code and have it execute inside Conway's Game of Life.
A demo shows Fable 5.2 bringing chalk art to life as an animated sequence, generated entirely in code.
A developer challenged Fable 5.2 to produce claymation-style animation and came away impressed by the results.
In pre-release testing, Fable 5.2 designed three functional electronic devices, including a home voice assistant hub, from a single paragraph prompt. An Anthropic team member reacted with "The future is now."
Fable 5.2 generated a browser-playable game demo featuring a boat ride through eight dreamlike worlds. The reaction: "bruhhhh."
The desktop app was incorrectly capping context at 200k instead of the full 1M. A fix is available now via app update.
In response to user feedback that MCP connectors in Claude.ai Settings apply globally across all projects, the product manager confirmed this will be addressed, with per-project trust boundary scoping on the way.
A developer complained that adding a project in the Claude Code desktop app took two minutes due to confusing navigation, and separately criticized the lack of easy bypass permissions. A team member acknowledged UI improvements are coming but pushed back on the bypass permissions ask, saying Auto Mode is intentionally designed to make elevated permissions harder to reach.
To help prevent distillation, preserved thinking is now enabled for all API accounts on Fable 5.2. Editing prior context in a conversation will return a 400 error; workarounds include compaction, context editing, or mid-conversation system messages. Claude Code, Managed Agents, and Agent SDK handle this automatically.
A new official guide explains when to use Fable 5.2, Opus 5.5, Sonnet, or Haiku, with advice on adjusting effort levels before switching models and routing subagents to smaller, cheaper models.
