Anthropic unveils Claude Fable 5.1, claiming up to 45% lower cost for agentic tasks

New Models and Pricing Claims

Anthropic announced the release of two upgraded AI models, Claude Fable 5.1 and Mythos 5.1, in early September 2026.

The company says Fable 5.1 delivers stronger performance than its predecessor while typically costing about 25 percent less and up to 45 percent less for complex agentic tasks.

The cost reduction is attributed to lower pricing on cached data that has already been processed and stored.

Mythos 5.1 is being introduced through the Project Glasswing program and is not yet broadly available.

Anthropic framed the upgrades as a response to customer concerns about price, data retention, and overly aggressive safety filters.

In a statement, Anthropic highlighted that Fable 5.1 includes “more precise safeguards” that are less likely to block routine biology queries.

By contrast, Mythos 5.1 retains the same biology restrictions that were present in the earlier version.

The firm also unveiled Enterprise Frontier Safeguards, which store user data on the customer’s own cloud servers rather than Anthropic’s infrastructure.

Anthropic described this approach as offering “complete privacy” and indicated a rollout beginning later this fall.

Early User Feedback and Performance

Early impressions from industry leaders began to appear shortly after the announcement.

Dan Shipper, founder of Every CEO, praised Fable 5.1 as “the strongest coding model we’ve used, but now it’s fast, token‑efficient, and crucially actually speaks like a normal person.”

Aaron Levie, CEO of Box, reported that his company’s AI agent using Fable 5.1 detected subtleties and ambiguities that the prior model missed in the same test.

Analyst Lisan al Gaib noted that Mythos 5.1’s low‑reasoning benchmark scores match those of its predecessor when set to maximum reasoning mode.

Anthropic said the new models are now permitted to assist in identifying software vulnerabilities.

However, the company will continue to route more specialized cybersecurity tasks—such as penetration testing, exploit generation, and binary‑based vulnerability scanning—to its Opus family of models.

Privacy, Safeguards, and Availability

Claude Fable 5.1 has been made available on all of Anthropic’s platforms, allowing developers and enterprises to integrate it immediately.

Access to Mythos 5.1 remains limited to participants in Project Glasswing, reflecting a staged rollout strategy.

The announced pricing advantage for agentic work could make large‑scale autonomous applications more affordable for businesses.

The privacy‑first data handling promised by Enterprise Frontier Safeguards may address regulatory pressures faced by enterprises handling sensitive information.

More precise safety filters aim to reduce false positives that previously interrupted legitimate queries, potentially improving user experience.

By keeping biology restrictions on Mythos 5.1, Anthropic signals a differentiated safety posture across its model portfolio.

The combination of cost savings, privacy guarantees, and refined safeguards positions Anthropic to compete more aggressively with other large‑scale AI providers.

Observers will watch how the reduced pricing on cached data influences overall usage patterns for autonomous agents.

The fall rollout of Enterprise Frontier Safeguards will provide a real‑world test of Anthropic’s “complete privacy” claim.

Why This Matters: Anthropic’s Fable 5.1 and Mythos 5.1 lower expenses for agentic tasks while enhancing privacy and safety controls.