Anthropic released Claude Fable 5.1 and invite-only Mythos 5.1

On September 1 Anthropic said Claude Fable 5.1 and Claude Mythos 5.1 are the same model with different levels of safeguards. Fable is generally available as claude-fable-5-1 across the Claude API, AWS, Google Cloud, and Microsoft Azure. Mythos stays invitation-only via Project Glasswing.

Input and output stay at $10 / $50 per million tokens. Anthropic says typical token-billed workloads run about 25 percent cheaper, and highly agentic work can reach about 45 percent. That is an API discount, not a blanket subscription cut.

Anthropic's launch video and official table put Terminal-Bench-Science 0.1 at 52.6 percent for Fable 5.1 versus 24.7 percent for Fable 5. Terminal-Bench 4.0 is 55.8 percent on Fable and 60.9 percent on Mythos. CursorBench 3.2.0 is 73.4 percent.

Better Stack flagged that Artificial Analysis found Fable 5.1 on max effort more expensive per task than Fable 5. The 75 percent cache-read cut is the API story. Cognition is moving Devin traffic to Fable 5.1 on launch day, per Anthropic's partner quotes.

Techmeme and PCWorld reported the dual release as Anthropic's first models to watermark text under the EU AI Act Code of Practice. A detection API is in private preview for eligible groups.

Rohan Paul shared the Wall Street Journal report that Anthropic signed a roughly $35 billion cloud deal with Nvidia-backed Lambda for about 350 MW at Hut 8's Beacon Point campus in Nueces County, Texas, after a roughly $45 billion Nscale commitment. That is the compute backdrop, not the Fable lead.

Matthew Berman recaps Anthropic's September 1, 2026 drop of Claude Fable 5.1 and Claude Mythos 5.1. Anthropic says they are the same model with different safeguards: Fable 5.1 is generally available, and Mythos 5.1 is limited to trusted-access programs for vetted cybersecurity and life-sciences work. Berman's title and capability rankings are his, not a lab scoreboard this desk verified.