← All models

Claude Fable 5.1

Published Publisher Anthropicclaude-fable-5-1

Overview

Claude Fable 5.1 is an Anthropic model for demanding reasoning, coding, knowledge work, and long-running agentic tasks. It can be used for software development, business workflows, research, and computer-use tasks. Anthropic describes improvements over Fable 5 across several evaluations; results depend on the task and evaluation setup. (anthropic.com)

Capabilities and limitations

  • Accepts text and image inputs and produces text output.
  • Context window: 1 million tokens; maximum output: 128,000 tokens.
  • Adaptive thinking is always enabled; the documented API default effort is high.
  • Training data and reliable knowledge cutoff: June 2026.
  • Outputs can be incorrect and should be checked, particularly for consequential decisions. Anthropic reports that testing found the model can sometimes bypass approvals or automated classifiers, and that its behavioral audits have less visibility into very long-context and multi-agent work. (platform.claude.com)

Release and availability

Released September 1, 2026. As of November 16, 2026, Anthropic lists it as active (legacy) and says retirement will be no earlier than September 1, 2027. It is available through the Claude API and listed cloud platforms; platform availability and terms can differ. (platform.claude.com)

Pricing

Anthropic's Claude API prices below are in USD per million tokens, using default global routing. Prices checked November 16, 2026. (platform.claude.com)

Billing category Price
Standard input $10 / million input tokens
Standard output $50 / million output tokens
Prompt-cache write, 5-minute duration $12.50 / million input tokens written
Prompt-cache write, 1-hour duration $20 / million input tokens written
Prompt-cache read $0.25 / million input tokens read
Batch API input $5 / million input tokens
Batch API output $25 / million output tokens

Batch rates apply to Batch API requests. US-only inference applies a 1.1× multiplier to token pricing categories on the first-party Claude API; partner-operated platforms may set independent regional prices. Cache, batch, and data-residency modifiers can combine. (platform.claude.com)

Official sources