Sarvam-30B

model
Sarvam
30B total / 2.4B active MoE (top-6 routing, GQA). 32K context. Trained from scratch for Indic languages. AA Intelligence: 12. Apache 2.0.
Blog PostArtificial Analysis

Model Details

Architecture MOE
Parameters 30B
Active params 2.4B
Context window 32,000
moeopen-weightmultilingual

Related