Sarvam releases Prism, an open-weight decision model that answers in 44 milliseconds
Sarvam AI has launched Sarvam Prism, a decision-making model distilled from its 105B base and built for structured, low-latency calls such as support routing, document triage and agent evaluation. The company reports a 44-millisecond average response on an Nvidia B200 and an 82.5 score on its Indian-language benchmark, with weights under Apache 2.0.
