Claude Fable 5

// Overview

Claude Fable 5 is Anthropic's most capable widely released model and the first member of its Mythos class — a capability tier above the Opus line — that the company made generally available. It launched on June 9, 2026 (API id claude-fable-5) on the Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, and Microsoft Foundry, built for the most demanding reasoning and long-horizon agentic work.

Claude Fable 5 shares the same underlying model as the restricted Claude Mythos 5, but ships with safety classifiers that can decline requests in high-risk areas (cybersecurity, biology and chemistry, and model-distillation attempts); refused requests instead route to Claude Opus 4.8. Anthropic reports that more than 95% of Fable 5 sessions involve no fallback at all. It carries a 1M-token context window and up to 128K output tokens, accepts text and image input, and runs adaptive thinking that is always on (extended thinking is not offered; the effort parameter controls thinking depth).

Three days after launch, on June 12, 2026, Anthropic suspended all access to Claude Fable 5 and Claude Mythos 5 to comply with a US government export-control directive — reportedly the first time the United States applied export controls to an AI model itself rather than the chips behind it. Anthropic publicly disagreed with the decision, which it tied to a narrow potential jailbreak, but disabled the models for all customers to remain compliant while working to restore access.

// Benchmarks

This model's scores

1. SWE-bench Verified95%
2. SWE-bench Pro80.3%
3. OSWorld-Verified (computer use)85%
4. Terminal-Bench 2.188%

Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.

// Pricing

5-minute cache write $12.50/MTok; 1-hour cache write $20/MTok; Batch API $5 input / $25 output per MTok. Same pricing for Mythos 5.

Pricing source ↗

// Strengths

• Frontier-level agentic coding — 95% on SWE-bench Verified and 80.3% on SWE-bench Pro, ahead of Claude Opus 4.8
• Strong computer use and terminal agents (85% OSWorld-Verified, 88% Terminal-Bench 2.1)
• 1M-token context window billed at standard per-token rates, with up to 128K output tokens
• Vision input strong enough to rebuild a web app's source from screenshots and read precise numbers from scientific figures
• Adaptive thinking always on, with an effort parameter to tune reasoning depth against cost

// Best for

• Reach for it for the most demanding multi-step agentic coding across large codebases when you want above-Opus capability.
• Use it for long-horizon agent workflows (computer use, terminal automation) where reliability matters more than per-token cost.
• Apply it to deep reasoning and analysis over very long documents that fit a 1M-token context.
• Use it for vision-heavy work like reconstructing UIs from screenshots or extracting figures from scientific PDFs.

// How to access

// Claude Fable / Mythos (Mythos-class) — every version

The full lineage of the Claude Fable / Mythos (Mythos-class) line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.