Anthropic Mythos 5 — The Frontier Tier Above Opus, Explained

Overview

Anthropic's Claude Mythos 5 is a frontier AI model, a tier above Opus, sharing its core capabilities with the public's Claude Fable 5 but with lifted safeguards. It represents the company's most capable and tightly controlled system, launched via a government program and briefly subject to export controls.

In this video

• Anthropic revealed two frontier models on June ninth, 2026: Claude Fable 5 for public use and Claude Mythos 5, the same underlying model with some of its safeguards lifted.
• Mythos-class models sit a full step above the Opus class, demonstrating state-of-the-art performance on nearly every benchmark, with a wider lead on longer and more complex tasks.
• Mythos 5 and Fable 5 share the same engine, featuring a 1 million-token context window and up to 128,000 output tokens.
• The models include persistent file-based memory, making them about three times better than Opus 4.8 and achieving 80.3% on the SWE-Bench Pro coding benchmark.
• The core difference between Mythos 5 and Fable 5 lies in their guardrails; Fable 5 defers sensitive prompts (e.g., cybersecurity, biology) in under 5% of sessions, while Mythos 5 answers directly.
• Mythos 5 was initially deployed narrowly through Project Glasswing, a program in collaboration with the United States government, providing controlled and limited access.
• Anthropic states that Mythos 5 possesses the strongest cybersecurity capabilities of any model in the world, capable of reading and reasoning about code and system weaknesses at an unmatched scale.
• Just three days after their June ninth launch, Mythos 5 and Fable 5 were subject to U.S. government export controls, leading to a two-and-a-half-week suspension of access.
• On June thirtieth, the export controls were lifted, and on July first, both models returned to users globally.
• The two-tier design, with a broadly available safeguarded version (Fable) and a gated, unlocked version (Mythos) for vetted partners, serves as a template for managing dual-use frontier capabilities.

Frequently asked questions

What is Anthropic Claude Mythos 5?

Claude Mythos 5 is Anthropic's most capable and tightly controlled frontier AI model, sharing its core with Fable 5 but with some safeguards lifted for direct responses to sensitive queries. It sits a tier above the Opus class models.

How does Claude Mythos 5 differ from Claude Fable 5?

Both Mythos 5 and Fable 5 share the same underlying powerful engine and capabilities. The key difference is that Fable 5 includes built-in safeguards, causing sensitive prompts to defer, while Mythos 5 has these safeguards lifted in certain areas, allowing it to answer directly.

What is Project Glasswing?

Project Glasswing is a program through which Anthropic first deployed Mythos 5 in collaboration with the United States government. It provided controlled, limited, and accountable access rather than an open public sign-up.

What are the main capabilities of Claude Mythos 5?

Claude Mythos 5 offers state-of-the-art performance across benchmarks, especially for complex tasks. Its headline capability is cybersecurity, demonstrating the strongest abilities in reading, reasoning about code, systems, and weaknesses at unmatched scale and speed.

Was Claude Mythos 5 subject to export controls?

Yes, Mythos 5, along with Fable 5, was briefly subjected to U.S. government export controls just three days after its June ninth launch, leading to a suspension of access for about two and a half weeks until the controls were lifted on June thirtieth.