Anthropic acquires Vercept
Terms were undisclosed; Vercept will wind down its own product, and Anthropic cited Claude's OSWorld computer-use score rising from under 15% in late 2024 to 72.5%.
- Money & business
- Labs & people
- Minor
Anthropic said it had acquired Vercept, a startup working on the perception and interaction problems involved in letting AI systems see and act within the software applications people use day to day. Financial terms were not disclosed, and Vercept said it would wind down its external product following the deal.
Anthropic framed the acquisition as reinforcing Claude’s computer-use capabilities, citing progress on the OSWorld benchmark — a test of AI systems operating a computer as a person would — where it said Claude models had improved from under 15% accuracy in late 2024 to 72.5%, approaching human-level performance on tasks such as spreadsheet navigation and web form completion. Vercept’s founding thesis, quoted by Anthropic, was that “making AI genuinely useful for completing complex tasks requires solving hard perception and interaction problems.”
The deal followed closely on the release of Claude Sonnet 4.6 earlier in the month, which Anthropic had also credited with a jump in computer-use performance, and fit a pattern among frontier labs of acquiring smaller specialist teams to absorb narrow technical capability into a general-purpose agent product rather than building every component in-house.