Anthropic launched Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026, introducing what the company calls the world's most advanced models for coding and knowledge work and an early glimpse of how AI systems will contribute to scientific research. The announcement, published on Anthropic's website, arrived alongside a set of pricing and data-handling changes that respond directly to customer feedback on cost, privacy, and safeguards.
The unusual two-name launch reflects a single underlying model with two safeguard configurations. Fable 5.1 is generally available to all users, while Mythos 5.1 is reserved for Anthropic's trusted access programs, where its safeguards are specifically designed to support work in cybersecurity and the life sciences. AWS confirmed the same day that Fable 5.1 is available on its platform in a launch post titled "Introducing Claude Fable 5.1 on AWS," and the news climbed to the top of Hacker News, drawing more than 600 points within hours of publication. For more context on this story, see our ongoing breaking AI news coverage.
Price Cuts and a Data-Retention Concession
The headline change for most customers will be cost. According to Anthropic, Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads wherever usage is billed by token. The reduction comes from lower pricing on cache reads — when the model reads inputs that have already been processed and stored. For highly agentic work, the company said savings will often be much larger, reaching up to approximately 45%.
The bigger structural change is a new system the company calls Enterprise Frontier Safeguards, or EFS. Anthropic says EFS gives enterprise customers complete privacy — equivalent to a zero data retention policy — while still maintaining state-of-the-art protection against adversarial use. The mechanism works by storing customer data in cloud infrastructure controlled entirely by the customer rather than by Anthropic. EFS will roll out to enterprise customers in phases beginning later this fall, and until it is generally available, eligible customers can run Fable 5.1 with zero data retention.
The privacy concessions follow visible customer pushback. CNBC reported on September 1 that Anthropic changed its data retention policy after objections from customers, and the EFS announcement is the company's most concrete response yet to enterprises that wanted frontier-model capability without their data resting on Anthropic's infrastructure.
Benchmark Gains Across Coding and Science
Anthropic's published evaluations show substantial jumps over its previous generation, though the numbers come from the company's own testing and should be read accordingly.
On Terminal-Bench-Science 0.1, an agentic scientific research benchmark, Fable 5.1 scored 52.6%, compared with 24.7% for Fable 5 in the same setup. Claude Opus 5 registered 29.0% and OpenAI's GPT-5.6 Sol scored 22.4% on the same benchmark. On Terminal-Bench 4.0, a terminal-based agentic coding test, Fable 5.1 scored 55.8% while Mythos 5.1, with its lighter-touch safeguards, reached 60.9% — up from Fable 5's 42.0%.
Knowledge-work and computer-use metrics improved as well. Fable 5.1 scored 1853 on the GDPval-AA v2 evaluation versus 1723 for Fable 5, and reached 60.9% on Humanity's Last Exam without tools and 65.0% with tools. On CursorBench 3.2.0, a coding benchmark, it scored 73.4% against Fable 5's 70.5%. In computer use, it scored 77.9% partial and 41.7% strict on OSWorld 2.0, and 31.4% on AutomationBench, up from 17.1% for its predecessor.
Anthropic also disclosed an important caveat: Fable 5.1 was evaluated with its production safeguards enabled, and on tasks where those safeguards intervened, the model scored zero — a factor the company said likely depresses its results on some benchmarks.
The company noted that effort settings matter for cost. Fable 5.1 defaults to High effort in Claude Code and Medium effort in Claude Cowork and on Claude.ai, and at lower effort levels it achieves results similar to or better than Fable 5 at a much lower cost.
Early Customers Report Productivity Gains
Anthropic published testimonials from early-access partners that emphasize both capability and economics. Cognition said it is moving its Opus 5 traffic in Devin to Fable 5.1 on launch day. "It matched or edged out Fable 5 in our testing at a lower cost per task, and with the new cache read pricing a Fable-class model is finally economical for the workloads we'd kept on Opus, starting with code review," said Walden Yan, Cognition's co-founder and chief product officer.
Quantitative trading firm Millennium offered a striking debugging anecdote. A senior portfolio manager said a rare crash — roughly one in a million runs — had gone unexplained by the firm's engineers for four to five years, and that every model he tried missed it. Fable 5.1 disassembled an external vendor library, matched it against the core dump, and traced the crash to a bug in that library.
Dan Shipper, CEO of the media company Every, said in his testing the model was about twice as fast as Opus 5 while using half as many tokens. And Ramp, the corporate card startup, described an unattended 38-hour run in which the model diagnosed a prior machine learning result as a label artifact, corrected it, kicked off six parallel experiments overnight, and returned with a result and next steps.
Softer Safeguards, With Strings Attached
On the safety side, Anthropic said its newest safeguards produce 60% fewer false positives in cybersecurity than before. The company also opened a capability shift: Fable 5.1 can now be used to discover software vulnerabilities, though not to develop exploits for them.
In biology, access to Mythos 5.1's advanced capabilities will run through a program developed in partnership with the US government, with enrollment for scientists expected to open soon. The company had already tightened biology-related protections on Fable 5 earlier this year, making the paired launch a continuation of that policy rather than a reversal.
Why It Matters
The launch lands in a market where inference prices have been falling fast — an AI Buzz Wire analysis recently found token prices hitting record lows — and Anthropic is clearly choosing to compete on price-performance for agentic workloads rather than defend premium rates. With published comparisons against OpenAI's GPT-5.6 Sol and its own Opus 5 line, the company is positioning Fable 5.1 as the default choice for long-running coding and research tasks. Whether independent evaluations and enterprise adoption confirm the pitch will become clear over the coming weeks.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
