Sarvam 30B — benchmark results

Sarvam AI's Apache-2.0 30B MoE (~2.4B active) trained from scratch in India under the IndiaAI Mission, strong in Indian languages and tool calling (February 2026). Provider: Sarvam AI. Released 2026-02-18. Access: API.

Unified ELO 1564 ± 14, rank #582 of 1776 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ZeroEval MATH-50097MATH-500 Score75.8
LLM Stats (HMMT25)74.2Score (%)41.7
ZeroEval GPQA Diamond66.5GPQA Diamond Score38.5
BenchLM40.7Overall Score15.7
LLM Stats (HMMT 2025)73.3Score (%)15.6
LLM Stats (BrowseComp)35.5Score (%)8.9
LLM Stats (Beyond AIME)58.3Score (%)0

Interactive version: theaggregate.ai/model?slug=sarvam-30b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.