Sage2 Launches: AI Can Update Judgments While Chatting, Responding in as Fast as 50 Milliseconds

Beating AI News Flash: US AI startup Levanto Labs has released Sage2, combining information extraction with real-time decision-making. It can automatically extract amounts, dates, user needs, and other information from chats, and continuously adjust its judgments based on new content.

For example, an AI customer service agent on a webpage can use Sage2 to continuously monitor user sentiment. A user initially only asks about a refund, then becomes increasingly impatient, and Sage2 can continuously update its judgment. When the user's level of dissatisfaction reaches a set threshold, the system can automatically transfer the user to a human agent without waiting for the user to ask.

The company says Sage2 takes about 150 milliseconds to extract information from short texts and as little as 50 milliseconds to update a judgment. The system retains conversation context and reuses computational caches, eliminating the need to resubmit the full chat history each time.

It is mainly aimed at scenarios such as customer service, business automation, and AI Agents, and is suitable for handling large numbers of repetitive judgment tasks. Compared with calling general-purpose large models, specialized decision models are expected to reduce latency and cost.

In the company's own evaluation, Sage2 scored 77.8 on information extraction accuracy, slightly higher than Cloudflare's data extraction model Clairic at 75.3.