Sarvam AI Launches High Speed Decision Model Prism For Enterprises

Sarvam Prism is built on Sarvam AI's 105B base model, aiming to automate fast, structured decisions in business workflows. The company states it delivers responses in 44 milliseconds on an NVIDIA B200 GPU, positioning it for real-time applications where quick processing is essential. Sarvam AI claims this speed, combined with an average score of 71.8 across three benchmarks, makes it the fastest among 25 decision models it tested that responded under 100 milliseconds.

For enterprises, Sarvam AI designed Prism to provide reliable confidence estimates, a feature intended to help businesses automate decisions while managing risk. In tests where answers were given with at least 90 percent confidence, the company reported error rates of 2.2 percent in English and 4.4 percent in Indian languages. Sarvam AI states these figures are below its own predicted error rates across various confidence thresholds.

A key feature of the launch is the release of Prism's weights under the Apache 2.0 licence. This allows enterprises to train the model on their own data and run it on their own infrastructure, offering control over data security and customisation. The model is also available through Sarvam's API, providing an alternative deployment method. Early applications include customer support routing, document intelligence, and security checks.