Sarvam Prism: a decision model built on Sarvam 105B (www.sarvam.ai)

Sarvam has launched Sarvam Prism, a cutting-edge decision model built on the Sarvam 105B architecture, aiming to revolutionize enterprise decision-making. This innovative model, inspired by Daniel Kahneman's System One thinking—characterized by quick and intuitive judgments—delivers unmatched performance with an accuracy score of 71.8 and a rapid response time of just 44 milliseconds on a B200 GPU. Prism outperforms 24 other decision models in a comprehensive benchmarking exercise, demonstrating that it achieves exceptional accuracy without sacrificing speed, crucial for enterprises relying on fast, reliable decision-making processes. This release is particularly significant for the AI/ML community as it introduces a highly calibrated model that enterprises can trust. Prism maintains error rates within its confidence bounds and is unaffected by the order in which options are presented, enhancing its reliability in real-world applications. Additionally, with open weights provided, businesses can customize and train Prism on their own datasets, further tailoring the model to specific workflows. This flexibility, coupled with high performance in various evaluations—including support for Indian languages and rigorous evaluation against established benchmarks—positions Prism as a transformative tool for decision-making in AI applications.