Models/Sarvam 30B

MDL

Sarvam 30B

Sarvam

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0.026 / $0.11

in / out

Context

—

Avg score

21.8

7 benchmarks

AA Intelligence

6.6

index

Speed

217 tok/s

TTFT 1.1s

Released

Mar 6, 2026

About

Sarvam 30B is a proprietary model in the Sarvam family from Sarvam. Benchmarked on 7 evals, averaging 21.8.

Benchmark scores · 7

avg 21.8 — higher is better

GPQA Diamondartificial-analysis · 2026-07

63.3

τ²-benchartificial-analysis · 2026-07

34.5

IFBenchartificial-analysis · 2026-07

26.5

SciCodeartificial-analysis · 2026-07

19.2

Humanity's Last Examartificial-analysis · 2026-07

Terminal-Bench Hardartificial-analysis · 2026-07

2.3

AA Long-Context Reasoningartificial-analysis · 2026-07

The research behind this model

all papers mentioning it →

BrahmicTokenizer-131K: An Indic-Capable Drop-In Replacement for o200k_baseMay 28, 2026

Mentions matched by name in title/abstract — from the arXiv + HF daily corpus.

Details

PublisherSarvam

FamilySarvam

Parameters—

License—

WeightsClosed

ReleasedMar 6, 2026

Performance & price

Output speed217 tok/s

Latency (TTFT)1.1s

First answer10s

Price in / out$0.026 / $0.11 per 1M

Blended price$0.047 / 1M

Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0.11/M