Sarvam 105B — benchmark results

Sarvam AI's 105B MoE (~9B active) sovereign Indian model, trained from scratch under the IndiaAI Mission with all 22 official Indian languages (February 2026). Provider: Sarvam AI. Released 2026-02-18. Access: Open.

Unified ELO 1650 ± 23, rank #367 of 1776 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ZeroEval MATH-50098.6MATH-500 Score96.8
LLM Stats (HMMT25)85.8Score (%)62.5
ZeroEval GPQA Diamond78.7GPQA Diamond Score59.7
LLM Stats (Beyond AIME)69.1Score (%)50
LLM Stats (HMMT 2025)85.8Score (%)37.5
LLM Stats (BrowseComp)49.5Score (%)23.2
BenchLM43Overall Score21.8

Interactive version: theaggregate.ai/model?slug=sarvam-105b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.