Sarvam AI launches Prism decision model with 44 millisecond responses

Sarvam AI launches Sarvam Prism, a decision model that produced responses in 44 milliseconds on an NVIDIA B200 GPU, according to ETCIO. Built on the Sarvam 105B base model, Prism handles structured tasks such as customer support routing, document analysis and AI agent evaluation.

Sarvam says Prism averaged 71.8 across three benchmarks, the highest score among 25 tested decision models that responded within 100 milliseconds. It scored 82.5 on the company's Indian language benchmark and 89 percent on an internal cybersecurity benchmark covering 5,517 questions.

The company released Prism's weights under the Apache 2.0 license, allowing businesses to customize the model with their own data and run it on private infrastructure. Prism is also available through Sarvam's API.