Sarvam 30B

high

Creator

Sarvam

Released

2026-03-06

Intelligence

6.6

Artificial Analysis Index

Coding

Artificial Analysis Index

In $/1M

$0.03

input tokens

Out $/1M

$0.11

output tokens

Blended $/1M

$0.05

3:1 blended

Speed

0

tokens / sec

Profile

License, openness, and modality — the governance layer.

License

Apache 2.0

commercial OK

Weights

Open

downloadable

Modalities

Parameters

30B

total

Sarvam-30B is an open-source 30B-parameter Mixture-of-Experts reasoning model from Sarvam AI trained from scratch and optimized for Indian languages, coding, and conversational workloads. It uses 128 sparse experts with 2.4B active parameters per token, Grouped Query Attention, and was pre-trained on 16 trillion tokens spanning code, mathematics, multilingual, and web data.

Capability profile

Category strength across 20 domains, via LLM Stats.

math
80legal
80finance
80language
80healthcare
80biology
70general
70physics
70
chemistry
70reasoning
70code
60writing
50creativity
50agents
40search
40frontend development
30
Via LLM Stats

Benchmark breakdown

Independent evaluation scores, normalized to 0–100.

GPQA Diamond
63
Humanity’s Last Exam
7
MMLU-Pro
SciCode
19
LiveCodeBench
MATH-500
AIME 2025
τ²-Bench (agentic)
35
Terminal-Bench Hard
2
IFBench
26
Via Artificial Analysis
© 2026 NYSGPT2525 LLC