Anthropic Unleashes Claude Fable 5: Enterprise AI’s New Powerhouse

Anthropic announced today the release of Claude Fable 5, its most capable model open to the public, alongside the restricted Claude Mythos 5.

Claude Fable 5 powers enterprise AI workloads

The new model builds on the Mythos‑class architecture but adds a fresh safeguard layer that redirects high‑risk queries to Claude Opus 4.8, preserving safety while delivering unmatched performance in code generation, document reasoning and vision tasks. In benchmark SWE‑bench Pro the model achieved ↑ 80.3% success, eclipsing GPT‑5.5’s ↓ 58.6%. Pricing is set at $10 per million input tokens and $50 per million output tokens – ↑ 50% lower than the previous Mythos Preview tier, yet still among the priciest offerings. Developers can access Claude Fable 5 via the Claude API (claude‑fable‑5) and through Anthropic’s web app, while Mythos 5 remains limited to vetted Project Glasswing participants and existing Mythos Preview users.

“Fable 5 compresses months of engineering into days,”

said a Stripe engineer, noting a full‑codebase migration completed in a single day on a 50‑million‑line Ruby repository. Other early adopters – Cursor, Replit, Figma – report breakthroughs in long‑horizon development and rapid prototyping. Beyond software, the model scored 1932 on GDPval‑AA and 29.8% on GDPpdf, indicating strong capability on dense PDFs, spreadsheets and charts, a boon for finance and legal teams still wrestling with legacy documents. The safety system flags requests touching cybersecurity, biology or chemistry, routing them to the less powerful Opus model; Anthropic claims over 1,000 hours of red‑team testing revealed no universal jailbreaks. A new 30‑day data‑retention policy applies to all traffic on Claude Fable 5 and Mythos 5, with Anthropic promising no use of retained data for model training. Enterprises must weigh the performance gains against the retention requirement, especially in regulated sectors. As AI races forward, Anthropic’s tiered rollout – a broadly available Fable 5 paired with a locked‑down Mythos 5 – may become the template for delivering frontier models without exposing the most dangerous capabilities. For context on how global crises shape AI adoption, see our recent analysis of the pandemic. Reuters covered the industry’s reaction to Anthropic’s approach.