Rate limits hit Anthropic users
Paying users reported hitting rate limits quickly during Mythos previews, a pattern some social posts read as evidence of compute ceilings under test conditions (x.com). Separately, Anthropic reportedly limited previews to about 12 partners and 40 organizations and referenced governance via a 'Long Term Benefit Trust' during the rollout (x.com).
Anthropic’s Mythos preview is reaching paying users through a narrow gate, and some of those users say that gate is closing fast with rate-limit messages. (anthropic.com) Anthropic said on April 7 that it would not make Mythos generally available and would instead use the model in a defensive cybersecurity program with a limited set of partners. Its Project Glasswing launch named 12 partners, including Amazon Web Services, Apple, Cisco, Google, JPMorganChase, Microsoft, and NVIDIA. (anthropic.com) Anthropic also said it extended access to “over 40 additional organizations” that build or maintain critical software infrastructure, alongside up to $100 million in Mythos Preview usage credits and $4 million in donations to open-source security groups. (anthropic.com) Mythos is the model behind that restricted rollout. Anthropic’s system card calls it the company’s “most capable frontier model to date” and says the increase in capability led Anthropic to keep it out of general release. (anthropic.com) The product question is not just who gets access, but how much access they get. Anthropic’s public materials describe a partner preview and a limited release, but they do not publish a public quota table for Mythos Preview in those rollout documents. (anthropic.com) That gap has left outside observers reading user reports as a signal about capacity as much as policy. TechCrunch reported on April 9 that some critics saw the restricted release as a way to steer top-end access toward large enterprise customers while also limiting opportunities for rivals to copy model behavior through distillation. (techcrunch.com) Anthropic’s stated reason for the narrow rollout is safety, not scarcity. In Project Glasswing, the company said Mythos Preview had already found thousands of high-severity vulnerabilities, including some in every major operating system and web browser. (anthropic.com) Axios reported on April 7 that Anthropic was withholding broader release because of concerns about the model’s ability to find and exploit security flaws, and described the preview as limited to a handpicked group of technology and cybersecurity organizations. (axios.com) The governance language around the rollout comes from Anthropic’s corporate structure, not a new body created for Mythos. The company said in September 2023 that its Long-Term Benefit Trust is an independent group of five financially disinterested members with authority that will grow to selecting and removing a majority of Anthropic’s board over time. (anthropic.com) Anthropic ties that trust to release decisions through its Responsible Scaling Policy. The company said changes to that policy must be approved by its board after consultations with the Long-Term Benefit Trust. (anthropic.com) So the current picture is two layers at once: Anthropic is openly saying Mythos is being held back for defensive-security testing, and users hitting limits are supplying a second, less formal read on how tightly the company is managing access while that test runs. (anthropic.com)
