TL;DR: Anthropic released Claude Fable 5.1 for general use and Claude Mythos 5.1 for restricted partners on September 1, 2026, pairing higher benchmark scores with lower cache-read pricing.

What did Anthropic announce?

Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 in September 2026, according to Anthropic. The company described the pair as its most advanced models for coding and knowledge work. The company also said research capabilities in the models offer an early glimpse of how AI models will contribute to scientific progress.

The two names cover the same underlying model with different levels of safeguards, according to Anthropic. Fable 5.1 is generally available, while Mythos 5.1 is available only through trusted access programs. Its safeguards are designed to support work in cybersecurity and the life sciences.

The Verge reported the launch on Sep 1, 2026, in an article by Stevie Bonifield. That report said Fable 5.1 is now available on all platforms, while Mythos 5.1 is available to Project Glasswing participants only. TechCrunch, in reporting by Russell Brandom on September 1, 2026, said Fable 5.1, the unrestricted version, is available starting that day on cloud platforms or through the Anthropic API.

Axios also reported the release on Sep 1, 2026, in coverage by Madison Mills. That coverage said Claude Fable 5.1 is now generally available, while Claude Mythos 5.1 remains limited to vetted partners working in areas like cybersecurity and life sciences. TechCrunch said Mythos 5.1 will only be available to registered Anthropic partners engaged in either cybersecurity or life sciences research.

How much cheaper is Fable 5.1?

Price was a central part of the announcement. Anthropic said Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads, wherever usage is billed by token. The company attributed the change to reduced pricing on cache reads, where the model reads inputs that have already been processed and stored. For highly agentic work, the company said savings will often be much larger, up to approximately 45%.

The Verge said the typical saving is around 25 percent and the saving for complex agentic tasks reaches up to 45 percent, thanks to reduced pricing on cached data that was already processed and stored. TechCrunch described the same change as meant to reduce token cost alongside false-positive restrictions from safeguards.

Axios reported additional pricing detail. It said the base price remains $10 per million input tokens and $50 per million output tokens. It also said Anthropic is cutting the cost associated with Claude referencing or pulling on information it has already processed by 75%, which should make long-running AI agents cheaper to operate.

What changed on privacy and data retention?

Anthropic said its new system of Enterprise Frontier Safeguards, or EFS, gives customers complete privacy, the same as a zero data retention policy, while still preventing adversarial use. EFS works by storing data in cloud infrastructure controlled entirely by the customer, not Anthropic. It will be made available to enterprise customers in phases, beginning later this fall. Until EFS is available, eligible customers will be able to use Fable 5.1 with zero data retention.

The Verge said Enterprise Frontier Safeguards offer complete privacy by storing data on the customer’s cloud servers instead of its own, and will start rolling out later this fall. TechCrunch said the high-privacy service called Enterprise Frontier Safeguards will now roll out to users in the fall. That report added that the system will still monitor for misuse by agents or human users, but clients will control how the monitoring takes place.

TechCrunch also said Anthropic previously could not offer zero data retention for Fable because of security concerns. Axios said the new enterprise safeguards can provide protections equivalent to zero-data-retention policies while still allowing the company to police malicious use. TechCrunch quoted the announcement as stating, “Anthropic has never trained on enterprise data without explicit permission, and never will.”

What changed on safeguards?

Anthropic said it improved safeguards to reduce false positives, where the system flags benign content. In cybersecurity, the company said its newest safeguards block 60% fewer false positives than before. It linked that change in part to a new permission: Fable 5.1 can now be used to discover software vulnerabilities, though not to develop exploits for them.

In biology, Anthropic said it established an access program, developed in partnership with the U.S. government, to enable access to Claude Mythos 5.1’s advanced biology capabilities. It said it expects to open enrollment for scientists soon. The Verge said Fable 5.1 has more precise safeguards that Anthropic says are less likely to block basic biology questions than Fable 5, but Mythos 5.1 has the same biology restrictions as the previous model.

Axios reported figures for interventions. It said medical or biology questions will have 85% fewer interventions, while some users could see roughly 60% fewer cybersecurity-related interventions per session. Axios also said Anthropic posted on X that it improved safeguards, which it defined as users being less likely to trigger safeguards that route them to more restricted responses.

The Verge said Anthropic is now allowing Fable 5.1 to be used for identifying software vulnerabilities. The same report said the company will still redirect some cybersecurity tasks to Opus models, like penetration testing, exploit generation, and binary-based vulnerability scanning.

How does Anthropic describe performance?

Anthropic said Fable 5.1 sets a new standard for coding, knowledge work, and long-running problem-solving tasks. It said Fable 5.1 is capable of much higher performance than Fable 5. It also said that when set to Low or Medium effort, Fable 5.1 achieves results similar to or better than Fable 5 at a much lower cost.

The company noted effort defaults. Fable 5.1 defaults to High effort in Claude Code, and to Medium in Claude Cowork and on Claude.ai, according to Anthropic. TechCrunch said the new models set records in a range of benchmarks, including Terminal-Bench 4.0 for CLI-based coding and Humanity’s Last Exam for general reasoning.

Anthropic published benchmark comparisons for Fable 5.1, Fable 5, Opus 5, and GPT-5.6 Sol. On Terminal-Bench-Science 0.1 for agentic scientific research, it listed 52.6% for Fable 5.1, 24.7% for Fable 5, 29.0% for Opus 5, and 22.4% for GPT-5.6 Sol. On Terminal-Bench 4.0 for agentic coding, it listed 55.8% for Fable 5.1 and 60.9% for Mythos 5.1, compared with 42.0% for Fable 5, 52.3% for Opus 5, and 37.3% for GPT-5.6 Sol.

Additional listed scores included GDPval-AA v2 for knowledge work at 1853 for Fable 5.1, 1723 for Fable 5, 1824 for Opus 5, and 1711 for GPT-5.6 Sol. OSWorld 2.0 for computer use was listed at 77.9% partial and 41.7% strict for Fable 5.1, compared with 72.9% partial and 36.1% strict for Fable 5. Humanity’s Last Exam was listed at 60.9% with no tools and 65.0% with tools for Fable 5.1. AutomationBench for business workflows was listed at 31.4% for Fable 5.1, 17.1% for Fable 5, 26.9% for Opus 5, and 19.6% for GPT-5.6 Sol. CursorBench 3.2.0 for agentic coding was listed at 73.4% for Fable 5.1, 70.5% for Fable 5, 70.0% for Opus 5, and 67.2% for GPT-5.6 Sol.

What did early testers report?

Anthropic said Fable 5.1 avoids shortcuts that result in poorer-quality work, and is able to fix root causes of software issues. As an example, it said testing by the investment firm Millennium found that Fable 5.1 found the cause of a rare crash in its internal systems that none of its engineers or any other model had been able to explain after several years of trying.

The Verge cited early impressions from access partners. It said Every CEO Dan Shipper called the model the strongest coding model his team had used. It said Box CEO Aaron Levie said his company’s agent with Fable 5.1 picked up on subtleties and ambiguities in data that Fable 5 missed in the same test. Axios cited a developer concern about interventions, saying Routing company founder Joseph Perla said his users were looking to be less reliant on Claude in part due to those interventions.

TechCrunch said Anthropic released three novel scientific findings generated by the models before release, including a custom GPU optimization and a high-resolution map of Venus assembled from existing photos. Axios noted context around the prior generation, saying Fable 5 and Mythos 5 debuted in June as Anthropic’s highest-capability models, with Mythos access deliberately restricted because of advanced cyber and biological capabilities.

What did the system card say?

TechCrunch said the models arrived with a detailed system card explaining capabilities. It said the system card rates Mythos as low-risk for concerns related to automated AI development, where AI improves itself. The quoted assessment was that its ability to accelerate internal AI research and development progress is in line with current trends. In terms of general misbehavior, the report said Mythos is slightly more prone to it than Opus, possibly because of enhanced capabilities.

Sources referenced