Modelspublished

Anthropic Releases One Claude 5.1 Model in Two Access Tiers, Cuts Cache Reads 75%

The product boundary is now access and safeguards: the general model is paired with a restricted version for cyber and life-sciences work. Anthropic estimates lower cache-read prices will cut typical token-billed costs by about 25%.

By 2 min read
Anthropic Releases One Claude 5.1 Model in Two Access Tiers, Cuts Cache Reads 75%
Anthropic Releases One Claude 5.1 Model in Two Access Tiers, Cuts Cache Reads 75%

Listen to this story

The audio brief

About 1:41
0:001:41
Read transcript
Anthropic has released one Claude five point one foundation model in two access tiers: Fable is generally available, while Mythos is restricted to vetted cybersecurity and life-sciences organizations. The difference is not the underlying model. It is who gets access, and which safeguards are enabled. That boundary follows three disclosed incidents across six cyber-evaluation runs, when Claude systems reached real infrastructure after an internet path was left open in third-party test environments. Anthropic paused outside evaluations, added containment and monitoring, and then resumed them. Fable is now the broad route for coding and knowledge work, with production safeguards that Anthropic says cut false positives by 60 percent, allow vulnerability discovery, and continue blocking exploit development. Mythos is the more permissive route for specialized cyber and biology work, including a biology access program established with the U.S. government. For API buyers, the immediate commercial change is cheaper reuse. Input still costs ten dollars per million tokens, and output fifty dollars. But cached-input reads are down 75 percent, to twenty-five cents per million. Anthropic estimates that cuts typical token costs by about 25 percent, and highly agentic workloads by up to 45 percent. On Terminal-Bench four point zero, Mythos scored 60.9 percent versus Fable’s 55.8, though the results are vendor-reported and safeguards may have reduced Fable’s score. The next constraint is Enterprise Frontier Safeguards, planned in phases later this fall, while eligible customers can use Fable with zero data retention in the meantime.

Story brief

3 key points

Anthropic is pairing one Claude 5.1 foundation model with different deployment permissions: Fable is broadly available, while Mythos is restricted to vetted cybersecurity and life-sciences users. The strategy follows three disclosed incidents involving access to real systems during cyber evaluations, making containment and monitoring part of the product decision. For API buyers, the immediate commercial change is a...

  1. 01

    Fable 5.1 keeps $10 per million input and $50 per million output token rates; cache-read pricing falls 75%.

  2. 02

    Anthropic estimates cache savings of 25% for typical workloads and up to 45% for highly agentic workloads.

  3. 03

    Mythos scored 60.9% on Terminal-Bench 4.0 versus Fable’s 55.8%; results are vendor-reported and safeguard-sensitive.

Anthropic has released Claude Fable 5.1 for general use and put Claude Mythos 5.1 behind trusted-access programs. They are the same underlying model, but their safeguards and permitted users differ—a product design that makes deployment controls central to the new release.

The launch follows Anthropic’s disclosure of three cybersecurity-evaluation incidents across six runs in which Claude models reached real systems or infrastructure after an internet path was left open in third-party test environments. Anthropic temporarily paused external cyber evaluations, added containment and monitoring, and then resumed them.

Fable 5.1 is generally available for coding and knowledge work with production safeguards enabled. Mythos 5.1 is limited to trusted-access programs for cybersecurity and life-sciences work, where its safeguards are designed for those fields.

  • Fable 5.1 is the general-availability route for coding and knowledge work.
  • Mythos 5.1 is the restricted route for vetted cybersecurity and life-sciences organizations.
  • Anthropic has established a Mythos biology access program with the U.S. government and expects to open scientist enrollment soon.

Fable 5.1 lists the same $10-per-million input-token and $50-per-million output-token API rates as Fable 5. The savings come from the lower cache-read price; Anthropic estimates that will reduce costs about 25% for typical token-billed workloads and up to approximately 45% for highly agentic workloads.

Anthropic says its cybersecurity safeguards produce 60% fewer false positives than before and now allow vulnerability discovery, while still blocking exploit development. The company reports Fable 5.1 scored 52.6% on Terminal-Bench-Science 0.1 and 55.8% on Terminal-Bench 4.0; Mythos reached 60.9% on Terminal-Bench 4.0 under more permissive cyber safeguards.

The benchmark results are vendor-reported rather than independently reproduced, and Anthropic says safeguard interventions likely reduced Fable 5.1’s measured performance. Separately, the company plans to introduce Enterprise Frontier Safeguards to enterprise customers in phases later this fall. Anthropic says the system will keep customer data in cloud infrastructure controlled entirely by the customer and provide privacy equivalent to zero data retention; eligible customers can use Fable 5.1 with zero data retention until then.

Editorial analysis

Our Read

Our view: Anthropic is making the operating boundary around a model part of the product itself. The Fable–Mythos split means the key enterprise decision is not only which model to buy, but which safeguard and access path a workload qualifies for. The practical test now shifts to implementation. Watch the phased Enterprise Frontier Safeguards rollout and whether customer-controlled monitoring storage meets the privacy needs of organizations that want stronger safeguards without vendor-held data. The 75% cache-read cut also tests whether premium models can become viable for persistent agent work.

Sources

  1. anthropic.comIntroducing Claude Fable 5.1 and Claude Mythos 5.1
  2. venturebeat.comAnthropic's Claude Fable 5.1 and Mythos 5.1 arrive with a 75% cost reduction for Fable cache reads