Claude Fable 5 — Anthropic's First Mythos-Class Frontier Model, Explained

Overview

Claude Fable 5 is Anthropic's first Mythos-class frontier model, establishing a new tier above Opus-class with state-of-the-art performance across benchmarks. It features a 1-million-token context window, file-based memory for improved output, and built-in safety fallbacks, making it ideal for complex coding, vision, and knowledge tasks.

In this video

• Claude Fable 5 is Anthropic's first Mythos-class model, a new capability tier that sits above the Opus class and is cleared for general use.
• The model is state-of-the-art on nearly every tested benchmark of AI capability, with its lead growing on longer and more complex tasks.
• Fable 5 launched on June 9, 2026, was briefly paused on June 12 under US export controls, and returned on July 1 once those controls were lifted.
• It pairs a 1-million-token context window with up to 128,000 output tokens per request, which is Anthropic's largest output window ever offered.
• One million tokens is roughly 750,000 words, about 3,000 pages, allowing an entire mid-sized codebase or a 1,000-document corpus to fit in a single call.
• Fable 5 is fully multimodal, handling text, image, and file inputs, with tool use, function calling, and the ability to read diagrams, charts, and tables nested inside PDFs.
• Persistent file-based memory allows Fable 5 to keep its own notes and use them to improve later output, performing about 3 times better than Opus 4.8 in testing.
• On SWE-Bench Pro, the hardest real-world software-engineering benchmark, Fable 5 posted 80.3%, the top score of any model tested, significantly higher than Opus 4.8's 69.2%.
• Fable 5 can rebuild a web app's source code from screenshots alone and achieved a leading 29.8% on a document-vision evaluation built from a dense economics PDF.
• Built-in safeguards automatically route prompts concerning cybersecurity, biology, chemistry, or health to Claude Opus 4.8, triggering in under 5% of sessions on average.

Frequently asked questions

What is Claude Fable 5?

Claude Fable 5 is Anthropic's most powerful generally available model, the first of its new Mythos-class, and represents a significant leap above the Opus-class in AI capability. It launched on June 9, 2026, and is state-of-the-art on nearly all tested benchmarks.

What is the context window size of Claude Fable 5?

Claude Fable 5 features a 1-million-token context window, which is roughly 750,000 words or about 3,000 pages, allowing for comprehensive single-call processing of large codebases or document corpuses. It also supports up to 128,000 output tokens per request.

How does Claude Fable 5 improve its work?

Claude Fable 5 utilizes persistent file-based memory, allowing it to keep notes as it works and use them to improve subsequent outputs. This memory helps it maintain focus across millions of tokens and enables self-correcting, durable performance for long tasks.

How does Claude Fable 5 perform on coding benchmarks?

On SWE-Bench Pro, the hardest real-world software-engineering benchmark, Claude Fable 5 achieved a top score of 80.3%. This is a substantial improvement over Opus 4.8's 69.2%, demonstrating its strength in complex, multi-step coding tasks.

Does Claude Fable 5 have safety features?

Yes, Fable 5 includes built-in safety safeguards; high-risk prompts related to cybersecurity, biology, chemistry, or health are automatically routed to the more conservative Claude Opus 4.8. This fallback mechanism triggers in under 5% of sessions on average, providing frontier capability with a safety net.