Claude Mythos 5

// Overview

Claude Mythos 5 is Anthropic's top-tier, Mythos-class frontier model, announced on June 9, 2026 alongside its public twin, Claude Fable 5. The two share the same underlying model and the same specs and pricing; the difference is safety scope. Claude Fable 5 ships with blocking safety classifiers that decline or fall back to Claude Opus 4.8 in sensitive domains (cybersecurity, biology, chemistry, model distillation), while Claude Mythos 5 carries those classifiers removed. Because of that, Mythos 5 is not generally available — it is offered only in limited release to approved customers through Anthropic's Project Glasswing, the company's program for hardening critical software, working with the US government and partners such as AWS, Apple, Cisco, CrowdStrike, Google, Microsoft, NVIDIA and Palo Alto Networks.

Under the hood, Claude Mythos 5 carries a 1M-token context window, up to 128K output tokens per request, and a January 2026 knowledge cutoff. It accepts text and image input (including diagrams, charts and tables nested in PDFs) and returns text. It is the successor to the invitation-only Claude Mythos Preview, which Anthropic credits with finding thousands of high-severity software vulnerabilities, and Anthropic describes the Mythos-class line as the strongest cybersecurity model in the world.

Claude Mythos 5 reaches the API as model ID claude-mythos-5, billed at $10 per million input tokens and $50 per million output tokens — about double Claude Opus 4.8 and roughly half the price of Mythos Preview. Days after launch, on June 12, 2026, Anthropic suspended access to both Claude Mythos 5 and Claude Fable 5 following a US export-control directive on access by foreign nationals; the model is currently unavailable pending resolution.

// Benchmarks

1. SWE-bench Verified95%
2. SWE-bench Pro (vendor-reported)80.3%
3. Hex core analytics benchmark90%

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

// Pricing

Shares Fable 5's pricing. 5-minute cache write $12.50/MTok; 1-hour cache write $20/MTok. Batch API: $5/MTok input, $25/MTok output. Limited availability via Project Glasswing.

Pricing source ↗

// Strengths

• State-of-the-art software engineering and agentic coding — 95.0% on SWE-bench Verified (Anthropic-reported)
• Long-horizon autonomy: plans across stages, delegates to sub-agents, and self-verifies, sustaining work across multi-day tasks
• Same frontier capability as Fable 5 but without the blocking safety classifiers, so requests in sensitive domains are not refused or routed away
• Strongest cybersecurity capability in the Claude line — built for vulnerability discovery and critical-software defense via Project Glasswing
• 1M-token context with vision over documents, diagrams, charts and tables embedded in PDFs

// Best for

• Defensive cybersecurity: finding and analyzing software vulnerabilities across large codebases (Project Glasswing partners)
• Long-horizon agentic engineering — multi-stage, multi-day autonomous coding and refactoring
• Complex knowledge work and data analysis with minimal human oversight
• Vision-grounded analysis of technical documents, scientific figures and PDFs
• High-autonomy research workflows for approved infrastructure and government partners

// 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.