Anthropic has completed training a new model: the successor to Mythos 5 is ready, but its fate is uncertain.
An intriguing development is brewing in the world of artificial intelligence. According to my information, Anthropic has completed training a new model that could become the direct successor to the acclaimed Mythos series. Insiders indicate that the system has already surpassed Mythos 5 in performance, but no official confirmation from the company has followed—neither a name nor stated specifications.
What is known about the new model?
According to analysis, the new system may be released under the name Mythos 5.1 or Mythos 6, or it might remain an internal development tool. This announcement came just nine days after U.S. export restrictions forced Anthropic to suspend work on the Mythos 5 and Fable 5 models. Recall that Fable 5 was intended for public access with enhanced security mechanisms, while Mythos 5 was used within the closed Glasswing project—the company's internal cybersecurity platform.
According to my information, about 50 partners using early versions of Mythos have already identified over 10,000 critical vulnerabilities in software. This speaks to the high efficiency of the lineup, even despite current restrictions.
Why hasn't development stopped?
On June 12, the U.S. Secretary of Commerce sent a directive to Anthropic CEO Dario Amodei, citing national security concerns. The restrictions affected all foreign nationals, including company employees born outside the U.S. Anthropic disabled both models worldwide, but as I see it, this has not halted progress.
Authorities found a way to bypass Fable 5's security, and Anthropic acknowledged the issue, emphasizing that such demands could block the release of new models across the industry. The company continues to push for the lifting of export restrictions, and rumors suggest that Amazon's warnings to officials played a key role here.
My expert opinion
Blocking development is unlikely to slow progress—rather, it could accelerate it by freeing up resources. Look at open-source competitors, such as GLM-5.2 from Z.ai: they handle programming just as well as much more expensive closed models. Whether Anthropic releases the new model publicly, restricts access through Glasswing, or keeps it for internal use, this will significantly shift the balance of power in the AI market. For now, the fate of Mythos 5's successor remains undecided, but one thing is clear: the race continues, and the stakes are high.