Sarvam Prism decision model hits 71.8 benchmark score at 44ms median latency
itsOmSarraf_ · x · 2026-11-03
Sarvam released Prism, a System One decision model built on Sarvam 105B for fast, accurate enterprise workflow decisions. It averages 71.8 across 3 decision benchmarks with a 44ms median latency — the best among decision models under 100ms — scores 82.5 on Indian-language evaluations, and shows 0% decision changes across 2,000 option-shuffled questions. Available via API and open weights (Apache 2.0).
Related event: India's Sarvam Open-Sources Prism, a 44ms Decision Model(3 posts)→
More from Models
- Why a chatbot spells 17 as "seveen": tokenization quirks exposed when reasoning is off — mmitchell_ai · 2026-11-04
- TheZvi diffs OpenAI's new Model Spec: philosophically similar, but new examples dodge hard cases — TheZvi · 2026-11-04
- LlamaIndex stacks 17 small optimizations for ~10x faster OCRia-1 token generation on a single B200 — llama_index · 2026-11-04
- Open-source 4B Smaug-Audio-Decisions tops audio decision leaderboard, beating Gemini APIs — bindureddy · 2026-11-04
- Reddit weighs 3 ChatGPT Plus accounts vs one Claude Pro upgrade (~€100 vs ~€120), EU dots caveat — Invalid-Function · 2026-11-04
- Reddit: Gemini 4 Argon reportedly near Claude and GPT-6 on new agent benchmark — Ok_Concept4102 · 2026-11-04