Anthropic, the AI safety company behind the Claude family of models, has been using an unpublished AI model internally referred to as 'Model 2,' according to a report by The Decoder.
What Happened
The Decoder reports that Anthropic is operating an internal model called 'Model 2' that has not been released publicly. Little technical detail about the model's capabilities or architecture has emerged, and the company has not formally announced its existence beyond these reported internal uses.
Why It Matters
The use of undisclosed models by leading AI labs raises questions about transparency in the industry. For developers and enterprises evaluating AI providers, understanding which models power various products can inform decision-making around reliability, safety standards, and long-term vendor relationships. The practice also highlights how frontier AI companies may be developing capabilities ahead of public releases.
The Bottom Line
Anthropic appears to operate internal model infrastructure beyond its publicly released offerings like Claude 3.5 Sonnet, though specific details about Model 2 remain limited based on available reporting.