Claude Fable 5.2

Model Specs

Item	Specification
Input	Text and images
Output	Text, with tool-call information supported
Context window	1M tokens
Maximum output	128K tokens
Vision	Supported
Function calling	Supported; forced tool choice is unsupported on the official API

Overview

Released by Anthropic on November 9, 2026, Claude Fable 5.2 targets demanding reasoning, extended agent sessions, and multi-step research. It is designed to carry analysis through to reviewable reports, spreadsheets, or presentations.

Anthropic reports improvements over Fable 5.1 in coding, reasoning, and vision, with output generation up to 30% faster. Software migration and scientific engineering examples demonstrate potential on difficult work, without guaranteeing success on every task.

Core Capabilities

• Complex software engineering: Analyze dependencies, plan migrations, and advance code changes.
• Deep reasoning: Address technical problems requiring staged analysis and verification.
• Long-horizon agents: Incorporate tool feedback throughout sustained tasks.
• Visual understanding: Analyze images alongside text.
• Research deliverables: Organize evidence into reviewable outputs.
• Scientific engineering assistance: Explore technical questions and proposals subject to domain validation.

Technical Strengths

Feature	Value
Million-token context	Accommodates extensive task material.
128K output limit	Provides room for lengthy analysis and generation.
Adaptive thinking	The model determines when and how to reason; thinking remains enabled.
Five effort levels	Supports low, medium, high, xhigh, and max; defaults to high.
Image and text input	Incorporates visual evidence into analysis.
Tool-use interface	Connects model decisions to application-provided execution.

FAQ

How should I choose between Claude Fable 5.2 and Claude Opus 5.5?

Evaluate Opus 5.5 first, then use Fable 5.2 for difficult tasks that still miss your acceptance criteria. Anthropic similarly recommends testing Opus at higher effort before upgrading.

Can I disable thinking in Claude Fable 5.2?

Adaptive thinking cannot be disabled, but effort controls how much processing the model applies. Establish a baseline at the default high level, then compare quality and duration at other settings on the same tasks.

How should I structure long-running tasks for Claude Fable 5.2?

Provide a clear goal, available tools, intermediate deliverables, and completion criteria. Save progress and validation records, requiring evidence at each stage rather than treating a lengthy response as proof of completion.

What should I check when migrating from Claude Fable 5.1 to 5.2?

Check conversation-history edits, thinking-block compatibility, computer-use tool versions, and advisor configuration. The official update identifies these four compatibility changes; validate multi-turn sessions before switching production traffic.

Why might a Claude Fable 5.2 tool request fail?

The official API does not support forced tool choice, and some older parameters and tool definitions require migration. Inspect request formatting, tool-selection settings, and error details instead of treating a model-name change as a complete migration.

How do I use the Claude Fable 5.2 API through LinkModel?

Open its model page on LinkModel and configure the published model identifier, authentication method, and request format. Validate a short text request first, then configure image input, tool calling, and reasoning options within the documented support scope.

How do I validate Claude Fable 5.2 on LinkModel?

Use representative business examples to check answer accuracy, tool execution, and complete-task duration. Include multi-turn conversations, long material, and recovery from failures, checking LinkModel’s documentation rather than assuming all Anthropic-specific headers and thinking-block options pass through.