Back
Technology

Anthropic Launches Claude Fable 5 and Claude Mythos 5

View source

Anthropic Launches Claude Fable 5 and Claude Mythos 5

A new generation of AI models arrives: state-of-the-art performance at half the price, with tiered safety controls.

Anthropic has released Claude Fable 5, a Mythos-class AI model, for general use, alongside Claude Mythos 5—a version with lifted safeguards reserved for approved cybersecurity partners.

Capabilities

Fable 5 is state-of-the-art on nearly all tested benchmarks, excelling in software engineering, knowledge work, vision, scientific research, and long-context tasks. It demonstrates improved performance over previous models, including Claude Opus 4.8, especially on complex, long-horizon tasks.

Safeguards

Fable 5 includes classifiers that detect potential misuse in cybersecurity, biology/chemistry, and distillation. When triggered, the model falls back to Opus 4.8. These safeguards are tuned conservatively, with less than 5% of sessions triggering a fallback.

Pricing

Both models are priced at $10 per million input tokens and $50 per million output tokens—less than half the price of Claude Mythos Preview.

Availability

  • Fable 5 is available today via the Claude API and on subscription plans temporarily.
  • Mythos 5 is initially available through Project Glasswing and will expand via a trusted access program.

Data Retention

Anthropic will retain 30-day data for Mythos-class models for safety purposes, with privacy protections and deletion after 30 days.

Performance Highlights

Fable 5 achieved top scores on several benchmarks including Hebbia’s Finance Benchmark, Cognition’s FrontierCode, and CursorBench. In a notable demonstration of its software engineering capabilities, the model performed a codebase-wide migration on a 50-million-line Ruby codebase in one day.

Early Feedback

Customers including Stripe, Cursor, GitHub, and others reported improvements in coding, reasoning, and analytics. Specific feedback highlighted autonomy, reliability, and efficiency gains.

Trusted Access Program

Anthropic plans to expand access to Mythos 5 for cybersecurity organizations and life science researchers through a trusted access program, scheduled to open soon.