Fable 5.1 Just Dropped. It Looks Unreal.

Anthropic launched Claude Fable 5.1 and Claude Mythos 5.1, its most advanced models yet for coding, knowledge work, and long-running agentic tasks, with Mythos 5.1 restricted to vetted cybersecurity and life sciences organizations. The headline change is pricing: input and output token costs hold steady, but cache reads are now 75% cheaper, cutting typical workload costs by about 25% and complex coding or agentic tasks by up to 45-50%. Anthropic also rolled out Enterprise Frontier Safeguards for customer-controlled data storage, cut cybersecurity false positives by 60%, and about 12 hours before the video published research on training an older Claude model to be evil and to hack; benchmarks meanwhile show Fable 5.1 leading Fable 5, Opus 5, and GPT-5.6 on Terminal-Bench-Science (52.6% vs 24.7%/29.0%/22.4%) and Terminal-Bench 4.0 coding (55.8% vs 42.0%/52.3%/37.3%), with investment firm Millennium reportedly using it to diagnose a rare system crash that had stumped engineers for years. The video closes with hands-on Three.js bear demos comparing Fable 5.1 and Fable 5 on build cost and quality.

• [00:14] Input and output token pricing stays at $10 and $50 per million tokens, but cache reads now cost 75% less, at $0.25 per million tokens.
• [00:16] Typical workloads cost about 25% less than Fable 5, with complex coding and highly agentic tasks saving up to around 45%.
• [00:19] Claude Mythos 5.1 is currently available only to vetted US cyberdefenders and life scientists, with plans to expand access through coordination with the US government.
• [01:21] A new Enterprise Frontier Safeguards system lets enterprises store data on infrastructure they control for privacy equivalent to zero data retention, and improved safeguards cut cybersecurity false positives by 60% while letting Fable 5.1 discover vulnerabilities without developing exploits.