Claude Fable 5.2 in Langdock

Claude Fable 5.2 is Anthropic’s model for demanding reasoning and long-horizon agentic work. In Langdock it is available with hosting in the EU. You can use it in chat, in agents and workflows, and via the API, where it costs €8.02 per 1 million input tokens and €40.08 per 1 million output tokens.

Written by LangdockUpdated

What is Claude Fable 5.2 good at?

Anthropic names these typical uses for Claude Fable 5.2[1]:

  • Agent sessions that run for hours
  • Deep research across many steps
  • Analysis carried through to a finished document, spreadsheet or deck

This is how Anthropic describes the model, not a test result of ours. Which tools the model can use in Langdock depends on the tools and integrations enabled in your workspace.

What are the key facts for Claude Fable 5.2 in Langdock?

Claude Fable 5.2: region, price and limits
FactValue
ProviderAnthropic[3]
Name in the model selectorFable 5.2[3]
Hosting region in LangdockEU[3]
Price per 1M input tokens€8.02[3]
Price per 1M output tokens€40.08[3]
Context window documented by Anthropic1M tokens[1]
Maximum output documented by Anthropic128K tokens[1]
Input types documented by AnthropicText and images[1]
Knowledge cutoff documented by AnthropicJune 2026[1]

The limits Anthropic documents apply to its own API; Langdock can set its own. The context window indicator next to the model selector shows the window size in use.

Token prices apply to the API, to model use in workflows and to extra usage, excluding VAT. They are converted from US dollars at the current exchange rate, so they can change slightly from day to day. In chat and agents, Core and Work seats with included models have no usage-based costs; pricing lists the plans.

When should you choose Claude Fable 5.2, and when another model?

Anthropic recommends starting with Claude Opus 5.5 for most workloads and moving to Claude Fable 5.2 for demanding reasoning and long-horizon agentic work, or when Opus 5.5 at a higher effort level still falls short. In its latest lineup of four models, Anthropic lists Fable 5.2 at the highest token price, so it is the choice for the hardest tasks rather than for everyday ones.[1][2]

The table compares what Langdock lists for Claude Fable 5.2 and for the alternatives. The model list has every model.

Claude Fable 5.2 and alternatives in Langdock
ModelProviderHosting regionInput / output per 1M tokens
Claude Fable 5.2AnthropicEU€8.02 / €40.08
Claude Opus 5.5AnthropicEU€4.01 / €20.04
Claude Sonnet 5.5AnthropicEU€2 / €10.02
GPT-6 AstraOpenAIEU€10.02 / €50.11

How do you use Claude Fable 5.2 in Langdock?

  • Chat: choose Fable 5.2 in the model selector of the chat input. You can switch models in the middle of a conversation and set a personal default model.
  • Agents: choose the model in the agent editor.
  • API: the Completion API has an endpoint that follows Anthropic’s own API format, with the region in the path. Its models endpoint lists the models available to your workspace in that region. The Agents API lists the models you can use for agents.
  • Admins: workspace admins enable or disable each model in the workspace settings and choose the default model for new chats and agents. If you do not see the model, ask your admin.

Details are in the model guide of the documentation.

Where is Claude Fable 5.2 hosted, and what happens to your data?

The Langdock model list shows the hosting region EU for Claude Fable 5.2[3]. In a new workspace only models hosted in the EU are enabled by default; admins decide which other models are switched on[8].

Check one thing before you enable Claude Fable 5.2: the Langdock documentation names the Fable models as models that can carry the “Extended data retention” label. For a model with this label, Anthropic stores input and output data for up to 30 days regardless of the data retention settings in your workspace, uses it exclusively for abuse prevention and not for model training, and deletes it automatically afterwards. According to the documentation this applies to Claude Fable 5.2 once the model shows the label in the model list of your workspace settings[8].

For all models, Langdock states that customer data is never used to train or improve AI models[9], and it publishes a data processing agreement (DPA)[10]. A hosting region alone does not make a use case GDPR-compliant: the guide to GDPR-compliant AI explains what Langdock provides and what remains your responsibility.

Frequently asked questions

Yes. The Langdock model list shows the hosting region EU for Claude Fable 5.2.

Via the API, in workflows and as extra usage, Claude Fable 5.2 costs €8.02 per 1 million input tokens and €40.08 per 1 million output tokens, excluding VAT. In chat and agents, Core and Work seats with included models have no usage-based costs.

Anthropic documents a context window of 1M tokens for its own API. Langdock can set its own limit; the context window indicator next to the model selector shows the size in use.

No. Langdock states that customer data is never used to train or improve AI models.

Sources

  1. Anthropic: Claude Fable 5.2 model page (checked on )
  2. Anthropic: Models overview (checked on )
  3. Langdock: Models with hosting region
  4. Langdock documentation: Model guide
  5. Langdock documentation: Agent configuration
  6. Langdock documentation: Completion API
  7. Langdock documentation: Models for the Agents API
  8. Langdock documentation: Workspace settings, models
  9. Langdock: Security
  10. Langdock: Data processing agreement (DPA)

Claude Fable 5.2 is a product of Anthropic. Product names are trademarks of their owners. Statements about the model were taken from Anthropic’s documentation on the date shown in the sources.