Anthropic intends to eventually offer Mythos, but it comes with some safeguards.
Anthropic announced Tuesday that it will be releasing a “Mythos-class” model called the Claude Fable 5, but the model has been given a different name because it has very important safety measures in place.
“Without safeguards, Fable 5’s capabilities in areas such as cybersecurity could be misused to cause significant harm,” Anthropic said in a statement.
The safeguards that some queries about certain topics will instead be ported to Anthropic’s Opus 4.8 model are initially tuned to be more conservative. Safeguards are triggered when Fable 5 detects a request related to cybersecurity, biology and chemistry, or distillation. In that case, Anthropic said users would be notified that their requests would be routed through Opus 4.8.
Anthropic is also continuing Project Glasswing, which it first announced in April, giving a limited number of technology, cybersecurity, and financial companies early access to the first version of Mythos, Claude Mythos Preview. The AI company said “a small group of cyber defenders and infrastructure providers” will have access to Claude Mythos 5, the same model as Fable 5, but “safety measures have been lifted in some areas.”
The company said Mythos 5 “has the strongest cybersecurity features of any model in the world,” and added, “Soon, we will expand access to Mythos 5 through a broader trusted access program.”
Anthropic said that even with its limitations, Fable 5 is “better than any model we’ve ever made generally available.”
“It is at the leading edge in nearly every tested benchmark of AI capabilities, demonstrating outstanding performance in software engineering, knowledge work, vision, scientific research, and many other areas,” the company said. “The longer and more complex the task, the greater the lead Fable 5 has over other models.”
Anthropic said it found that Stripe reduced the engineering effort for Fable 5 to a few months. The company said that when it comes to financial tasks, IMC found that Fable 5 “meets nearly all transactional analysis assessments, including fact-finding, conceptual reasoning, root cause analysis, and expected value analysis.”
Like previous Anthropic models, Fable fits into the theme of naming AI models based on their composition format. The AI startup said Fable was chosen because it shares similarities with Mythos.
“Fable comes from the Latin word fabula, ‘that which is told,’ which is similar to Greek mythology,” the company said.
The release comes at a critical time for Anthropic, which recently announced it had secretly filed a draft S-1 and officially began its march toward a blockbuster IPO.
Anthropic’s biggest rival, OpenAI, also announced on Monday that it had also confidentially filed an S-1. SpaceX, which includes Elon Musk’s xAI, is scheduled to begin trading and go public on the Nasdaq on Friday.
Anthropic’s initial announcement of Mythos surprised financial markets and governments around the world, raising concerns that AI models had advanced to the point where they could discover critical vulnerabilities in software and cybersecurity.
President Donald Trump recently signed an executive order that allows AI companies to voluntarily grant the federal government access to their Frontier AI models up to 30 days before release. The order specifically prohibits the government from imposing mandatory reviews on AI companies.