Home / Models / Claude Fable 5.1

Claude Fable 5.1

Anthropic · Multimodal model

Claude Fable 5.1 is Anthropic's highest tier model, released on September 1, 2026 together with Claude Mythos 5.1, which has the same weights but different safeguards. It is aimed at hard reasoning and long-running agent tasks and costs more than…

Current
ReleasedSep 1, 2026
Context window1M tokens (128K max output)
Input → outputtext, image → text
API price$10 / input MTok, $50 / output MTok
Consumer accessEnterprise; Max and Team premium seats (up to 50% of weekly limits); Pro via usage credits; not on Free

Best for

The most demanding reasoning and long-horizon agentic work where Opus 5.5 still falls short.

Overview

Claude Fable 5.1 is Anthropic's highest tier model, released on September 1, 2026 together with Claude Mythos 5.1, which has the same weights but different safeguards. It is aimed at hard reasoning and long-running agent tasks and costs more than the Opus line.

Strengths and limits

  • Anthropic's top-tier model for demanding reasoning and long-running agents
  • Strong on science and knowledge-work evaluations (Terminal-Bench-Science, HLE)
  • Cache reads cut 75% to $0.25/MTok, lowering agentic workload costs
  • Slowest and most expensive current Claude model
  • Dual-use cyber and life-science requests are redirected to Opus models; full access needs Mythos 5.1 via vetted programs

Published benchmark scores

Terminal-Bench 4.055.8% Source: www.anthropic.com
Humanity's Last Exam (with tools)65.0% Source: www.anthropic.com
OSWorld 2.0 (partial)77.9% Source: www.anthropic.com
Terminal-Bench-Science 0.152.6% Source: www.anthropic.com

Scores are as published by the source; settings (reasoning effort, tools, attempts) differ between labs, so compare with care. How to read benchmarks →

Sources

Facts checked Oct 14, 2026
  1. platform.claude.com/docs/en/about-claude/models/overview
  2. anthropic.com/claude-fable-and-mythos-5-1
  3. claude.com/pricing