Summary
Anthropic has released Claude Fable 5.1 and Mythos 5.1, claiming them as the world’s most capable models for coding and knowledge work. In this first-look video, Nate Herk walks through the announcement blog and runs quick live comparisons between Fable 5 and 5.1 on his Claude desktop app, testing the models against an identical prompt asking each to generate an animated 3D cartoon bear riding a bike.
The headline claim from Anthropic is roughly 25% lower costs for typical workflows — and up to 50% for highly agentic workloads — without changing list pricing of $10 per million input tokens and $50 per million output tokens. The savings come primarily from reduced pricing on cache reads, where previously processed prompt inputs are stored and reused across turns. Anthropic is also launching EFS (Enterprise Frontier Safeguards) alongside a zero data retention option, storing data in customer-controlled cloud infrastructure — a move aimed at regulated industries that previously could not adopt Fable 5 due to data residency concerns.
Mythos 5.1 mirrors Fable 5.1 in capability but ships with more permissive safeguards for vetted users in cybersecurity and life sciences, available through Anthropic’s cyber verification and life sciences verification programs. Herk notes that 5.1’s demo output — which spontaneously named its bear character Barnaby — showed noticeably improved design quality and live artifact rendering compared to Fable 5. A deeper performance review is promised after extended hands-on testing.
📺 Source: Nate Herk | AI Automation · Published September 01, 2026
🏷️ Format: News Analysis
