Anthropic just shipped two new models, Claude Fable 5.1 and Mythos 5.1, and the headline is money. According to The Verge AI, the company built these releases to answer three specific complaints from customers: price, data retention, and safeguards that block too much. Fable 5.1 is the star of the launch, and Anthropic claims it beats the previous Fable 5 on performance while costing less to run.
Here’s what Anthropic is putting on the table.
1. Cheaper agentic work, by a wide margin
Anthropic says Fable 5.1 costs around 25 percent less than Fable 5 for typical use, and up to 45 percent less for complex agentic tasks. The Verge AI reports the savings come from reduced pricing on cached data, meaning information that’s already been processed and stored. For teams running agents that chew through the same context repeatedly, that’s where the real cost lives.
2. Faster and more natural output
Early access users are talking. Every CEO Dan Shipper called it “the strongest coding model we’ve used, but now it’s fast, token-efficient, and crucially actually speaks like a normal person.” That last part matters more than it sounds. A model that writes like a human instead of a corporate memo is easier to drop into real products.
3. Sharper reasoning on ambiguous data
Box CEO Aaron Levie said his company’s agent running Fable 5.1 caught subtleties and ambiguities in data that Fable 5 missed on the same test. On the benchmark side, one observer noted that Mythos 5.1 on low reasoning scores the same as its predecessor cranked to Max reasoning. If that holds up, you get the same quality for far less compute.
4. More precise safeguards
Fable 5.1 ships with what Anthropic calls “more precise safeguards,” and the company says it’s less likely to block basic biology questions than Fable 5. Worth noting: Mythos 5.1 keeps the same biology restrictions as the older model, so the loosening isn’t across the board.
5. Data retention on your own servers
Anthropic addressed the privacy complaint with Enterprise Frontier Safeguards, which it says offer “complete privacy” by storing data on the customer’s cloud servers instead of Anthropic’s. That rollout starts later this fall. For regulated industries that can’t send data off-premises, this is the kind of change that unlocks a purchase.
6. Security work, with limits
Anthropic is “now allowing Fable 5.1 to be used for identifying software vulnerabilities.” It’s not a full green light, though. The company will still redirect certain cybersecurity tasks to its Opus models, including penetration testing, exploit generation, and binary-based vulnerability scanning.
Availability
Claude Fable 5.1 is live now on all platforms. Mythos 5.1 is locked to Project Glasswing participants only, so most people won’t touch it yet.
Why this matters
What stands out here is the target. Anthropic isn’t chasing a flashy benchmark number for the press cycle. It’s fixing the practical friction that keeps businesses from deploying agents at scale: cost, privacy, and models that refuse too many reasonable requests. The 45 percent cut on cached agentic work is a direct shot at the biggest hidden expense in production AI, and moving data storage onto customer servers removes a common blocker for enterprise deals.
The pattern across the industry is clear. The competition has shifted from raw capability to who makes AI cheap and safe enough to actually run in production. Anthropic is betting that better economics and enterprise-grade privacy win more customers than another point on a leaderboard. Whether the cost claims hold up under real workloads is the thing to watch next.
For the full breakdown and early reactions, check the original report at The Verge AI.