Claude Fable 5.1

Coding, debugging, and long-horizon reasoning for agents.

Playground
History
APIPricing
~6.89s

Chat

0 messages

Press Enter to send, Shift + Enter for new line • Max 5 images (10MB each)

Claude Fable 5.1 — Text Generation (LLM)

What is Claude Fable 5.1?

Claude Fable 5.1 is Anthropic's most capable widely available model, built for demanding reasoning, coding, and long-horizon agentic work. It always reasons before answering, ships with a 1M-token context window and 128k maximum output, and exposes a selectable reasoning-effort dial so you can trade cost for depth on each request. Through the API you send a prompt (plus an optional image for multimodal analysis and an optional system instruction) and receive plain text or schema-validated JSON.

Key Features

  • •Selectable reasoning effort: low, medium, high, xhigh, max.
  • •1M-token context window and 128k max output tokens.
  • •Multimodal image understanding for charts, tables, diagrams, and PDFs.
  • •JSON-Schema structured output via response_format.
  • •Optional system instruction for role, tone, and format control.
  • •Tuned for agentic coding, code review, and multi-file changes.

Best Use Cases

Coding is the headline strength: implement features that span a codebase, review changes, chase down rare bugs, and do performance work — with the model writing its own tests to verify a fix. In our testing it took a buggy merge_intervals function, named the exact defective line, explained the edge case it broke on, and returned a corrected, self-tested version. Beyond code it excels at multi-step reasoning, deep research, structured data extraction, and long-context document analysis across finance, legal, and analytics workflows.

Prompt Tips and Output Quality

Be concrete and give the model the full problem, including any code or data, inline. Raise effort to high or xhigh for coding and multi-step analysis, and max when correctness outweighs latency; keep it low for chat, extraction, and classification. Ask for tests, a checklist, or a schema so outputs are verifiable, and set a system instruction to lock the output format.

FAQs

What is Claude Fable 5.1 best at? Ambitious, long-running coding, reasoning, and knowledge work.

What does the effort parameter do? It controls reasoning depth; higher levels reason harder and cost more output tokens.

Does it accept images? Yes — it reads charts, tables, and diagrams alongside your prompt.

How large is the context? 1M input tokens with up to 128k output tokens.

Can it return JSON? Yes, via a response_format JSON Schema.