Claude Fable 5.1 Is Now Available in Buda
Buda now uses Fable 5.1 for difficult Agent work, with paid access and four reasoning levels.
Claude Fable 5.1 Is Now Available in Buda
Buda now uses Claude Fable 5.1 for its highest-tier Claude model. It is available to workspaces with an active paid subscription and is designed for difficult Agent work that benefits from longer planning, deeper reasoning, and repeated tool use.
This is a direct model replacement, not an alias. The previous claude-fable-5 model ID is removed from Buda. Model API users should switch to claude-fable-5-1.
What changes in Buda
Four product changes matter.
First, the model selector now shows Claude Fable 5.1. Existing Agent sessions refresh their model options so they do not keep an outdated name or multiplier.
Second, Fable 5.1 is a subscription-only Premium model. Free workspaces cannot select it.
Third, Buda displays a 5.0x Credit multiplier, and the actual Credit deduction is aligned with that same relative rate against the Claude Sonnet baseline. It is not a fixed price per request; actual usage still depends on tokens, cache, tools, and minimum-charge rules.
Fourth, Buda offers Low, Medium, High, and xHigh reasoning levels for Fable 5.1.
Choose the effort level from the task
Use the lowest effort level that reliably meets the acceptance criteria.
- Low: bounded work with a clear format and a quick factual check.
- Medium: normal multi-step work involving files, planning, or tools.
- High: difficult work with several dependencies or expensive mistakes.
- xHigh: the hardest cases where additional reasoning is worth more time and Credits.
Higher effort does not grant more permission or remove the need for review. Customer messages, production changes, payments, deletion, and publication should still stop for a person to inspect.
When Fable 5.1 is worth the premium
Anthropic positions Fable 5.1 for demanding reasoning and long-horizon Agent work. In Buda, that makes it a sensible escalation model when:
- an Agent must investigate a large codebase, implement a change, and verify it;
- a task combines many documents, files, and tool calls;
- work continues for hours and needs visible progress and recovery;
- the result is an editable document, spreadsheet, presentation, or website;
- failed attempts and reviewer corrections cost more than the model difference.
Routine extraction, tagging, and formatting usually belong on a lower-cost model. A premium model is useful when it changes the acceptance rate, not merely because it is available.
The model ID change is intentional
Buda does not silently map claude-fable-5 to the new model. Integrations that select a model through the Buda Model API must send claude-fable-5-1.
This keeps an explicit contract: the model requested by an integration is the model Buda routes. Update stored presets, scripts, and API clients before relying on the new version.
For the three upstream migration boundaries involving forced tools and thinking blocks, use the Claude Fable 5.1 upgrade guide.
A stronger model still needs a managed workspace
Model capability is only one layer of an Agent system. The Agent also needs the correct context, tools, permissions, stopping conditions, and review points.
In Buda, Fable 5.1 can work with persistent files and real tools while people keep the task, execution, and deliverable visible. The division remains clear: the model handles difficult execution; people decide whether the result is acceptable.
Select Claude Fable 5.1 from an Agent's model menu in a paid Buda workspace. Current model multipliers are listed on the Buda pricing page.
FAQ
Can Free workspaces use Claude Fable 5.1?
No. It requires an active paid Buda subscription.
What does 5.0x Credits mean?
It is Buda's relative model multiplier against the Claude Sonnet baseline. It does not mean every request consumes the same number of Credits.
Does the old claude-fable-5 ID still work?
No compatibility mapping is included. Model API callers should use claude-fable-5-1.
Which reasoning level should I start with?
Start with Medium for normal multi-step Agent work, then move down or up based on measured acceptance, latency, and Credit use.