Anthropic Quietly Tests Mythos, Its Most Powerful AI Model Yet
Anthropic is testing 'Mythos," a previously undisclosed and highly-capable AI model, the company said after accidental data leak reveals model's existence
Exclusive: Anthropic is testing 'Mythos' its 'most powerful AI model ever developed'
Anthropic, the San Francisco-based artificial intelligence company, has confirmed the existence of a previously undisclosed AI model codenamed "Mythos" after an accidental data leak brought the project to public attention earlier this week. The company, which is backed by billions in funding from major investors including Google and Salesforce, described Mythos as its "most powerful AI model ever developed" in a brief statement released late Wednesday evening. The acknowledgment came after internal documents referencing the model were inadvertently published to a public-facing server, prompting widespread speculation across social media and technology forums.
According to sources familiar with the matter, Mythos has been in development for several months and represents a significant leap in capability beyond Anthropic's current flagship Claude model series. The leaked documents suggest the model demonstrates remarkable improvements in reasoning, code generation, and multi-step problem solving, though the full extent of its capabilities remains unclear. Anthropic has not yet disclosed the model's architecture, parameter count, or the datasets used in its training, citing ongoing safety evaluations as the reason for its secrecy.
The accidental disclosure has reignited debate within the AI industry about transparency and the pace of frontier model development. Competitors including OpenAI, Google DeepMind, and Meta have all been racing to develop increasingly powerful systems, and the revelation of Mythos suggests Anthropic may be further ahead in that race than previously known. AI safety researchers have expressed both excitement and concern, noting that the company's decision to keep the model under wraps may reflect the seriousness of the safety testing it requires before any public release.
Anthropic said in its statement that Mythos is currently undergoing rigorous internal and third-party safety evaluations and that the company has no immediate plans for a public launch. "We are committed to ensuring that our most capable systems meet the highest standards of safety and reliability before any broader deployment," a company spokesperson said. The firm added that it would share more details about Mythos "in the coming weeks" and pledged to work with policymakers and independent evaluators as it determines the appropriate path forward for the model.