An Anthropic model for demanding reasoning, long-horizon agentic work, and text-and-image analysis with adaptive thinking.
Use anthropic/claude-fable-5-2 through the sjolt LLM API.
anthropic/claude-fable-5-2Token rates
anthropic/claude-fable-5-2Similar Models
An Anthropic model for careful coding, long-document analysis, and visual understanding with optional adaptive thinking.
An Anthropic model for coding, long-context knowledge work, and visual understanding with adaptive thinking.
An Anthropic model for responsive coding, long-document analysis, and text-and-image understanding with adaptive thinking.
Model overview
Carry complex technical work from investigation to a finished result with adaptive thinking through the native Anthropic Messages API.
Provide the relevant code, constraints, tools, and acceptance criteria. Keep the conversation append-only and return complete native assistant blocks when continuing tool workflows.
Use up to 1,000,000 context tokens for repositories, source documents, tool results, and conversation history. Label sources and request explicit support for conclusions.
Combine images with precise questions about diagrams, screenshots, charts, or physical systems. Responses are text; the illustrations on this page are conceptual artwork.
Model characteristics
Up to 1,000,000 context tokens and 128,000 output tokens. max_tokens includes thinking and the final response.
Thinking stays enabled. Select low, medium, high, xhigh, max through output_config.effort; the default is high.
Send model, messages, and max_tokens to /v1/messages, with system instructions in the top-level system field. Use automatic tool selection and preserve signed thinking blocks in append-only tool conversations.
Describe the objective, evidence, constraints, available tools, and finish condition. Use lower effort for bounded tasks and raise it only when your own evaluations show a benefit.
FAQ
Send model anthropic/claude-fable-5-2 to the Anthropic Messages endpoint /v1/messages with messages and max_tokens. Public API requests can use native streaming.
No. Omit thinking or set thinking.type to adaptive, then choose low, medium, high, xhigh, max through output_config.effort. The default effort is high.
Set max_tokens to a positive integer up to 128,000. Thinking and final output share this allowance within the 1,000,000-token context window.
Provide text and images through native Messages content blocks. The model returns text, and the Playground accepts image attachments for follow-up conversations.
Keep system instructions and tool definitions stable, append new turns without editing earlier content, and return every assistant block exactly as received. Use automatic tool selection rather than forcing a specific tool.