Anthropic publicly launched upgraded versions of its advanced AI models, Claude Fable 5.1 and Mythos 5.1, on Tuesday, September 1, 2026 [1, 2, 3, 4, 5]. Fable 5.1 is now generally available on cloud platforms and through the Anthropic API. Mythos 5.1 remains restricted to trusted partners in cybersecurity and life sciences due to compliance controls [1, 2, 3, 4, 5].
Claude Fable 5.1 outperforms its predecessor on key benchmarks such as the Terminal-Bench 4.0 coding and scientific research tests, scoring 52.6% compared to Fable 5's 24.7% on science benchmarks [3]. The model excels particularly at complex coding and problem-solving tasks, receiving praise from users. Aaron Levie, CEO of Box, said, "Our agent with Fable 5.1 picked up on subtleties and ambiguities in data that Fable 5 missed in the same test" [2]. Dan Shipper, CEO of Every, described it as "the strongest coding model we’ve used, but now it’s fast, token-efficient, and crucially actually speaks like a normal person" [2].
Operational costs have dropped substantially with Fable 5.1. Pricing for input and output tokens remains at $10 and $50 per million tokens respectively, but cache read prices have decreased by 75% to $0.25 per million tokens [3, 2, 4]. This has reduced general workload costs by approximately 25% and up to 45% for complex agentic tasks, lowering expenses for usage-intensive applications [2, 3, 4, 5].
Mythos 5.1 maintains the same strict biology-related restrictions as its previous version to ensure compliance in sensitive fields. Fable 5.1 adds more precise safeguards that reduce false positives on benign biology queries while maintaining safety controls [1, 2, 3, 4]. However, Fable 5.1 is slightly more prone to misaligned behavior than Anthropic’s Opus 5 model, despite improved behavior compared to earlier Anthropic systems [1].
Anthropic also announced the upcoming fall 2026 rollout of its Enterprise Frontier Safeguards privacy system. This zero data retention architecture allows customers to store data on their own cloud servers. Promising no enterprise data training without explicit permission, Anthropic stated, "Anthropic has never trained on enterprise data without explicit permission, and never will" [1, 2, 4].
The releases position Anthropic’s models for increased adoption in coding, scientific research, cybersecurity, and life sciences while offering significant cost savings and enhanced privacy controls. The Enterprise Frontier Safeguards launch scheduled for fall 2026 will expand secure deployment options.
