How V7 gives AI agents institutional memory

OpenAI highlighted how startup V7 is using its models to give AI agents a more durable sense of company context. The company’s V7 Go platform organizes information from documents, data rooms, spreadsheets, emails, and internal tools into a Context Graph that agents can query while working on complex enterprise tasks. V7 said the system helps finance, insurance, and real estate teams handle long workflows more accurately. According to the article, GPT5.6 Luna is used for largescale information extraction, GPT5.6 Terra and Sol support reasoning and tool use, and GPT6 Astra is being tested on the most difficult graphquery tasks. OpenAI said V7 reported 89% accuracy for Astra on its hardest benchmark and lower cost per document with Luna. The post also says V7 is using this shared context to speed up tasks such as deal screening and underwriting, while keeping an auditable trail of decisions. The company’s longerterm goal is to make workflows more proactive by flagging inconsistencies when underlying facts change.