Skip to main content
AITier

Sarvam 105B (high)

SarvamOpen WeightApache 2.0 · Commercial OK

Description

Sarvam-105B is Sarvam AI's flagship open-source Mixture-of-Experts reasoning model built for complex reasoning, coding, and agentic workflows. It uses 128 sparse experts with Multi-head Latent Attention for efficient long-context inference and was pre-trained on 12 trillion tokens spanning code, mathematics, multilingual, and web data.

Release Date
2026-03-06
Parameters
105.0B
Context Length
131K
Modalities
text

Capability Radar

12
general
26
coding
74
reasoning
44
scienceest.
50
agents
0
multimodal

Science uses a reasoning proxy when dedicated science benchmarks are unavailable.

Rankings

Domain#RankScoreSource
Agentic Capability87
50.0
LS
Code Ranking498
1.0
AA
General Ranking380
29.0
AA
Science222
46.0
AA

Benchmark Scores (LLM Stats)

Agents

BrowseComp49.5%SR

Biology

GPQA78.7%SR

Code

SWE-Bench Verified45.0%SR

Creativity

Arena-Hard v271.0%SR

Finance

MMLU90.6%SR
MMLU-Pro81.7%SR

General

IFEval84.8%SR
LiveCodeBench v671.7%SR

Math

MATH-50098.6%SR
AIME 202596.7%SR
HMMT2585.8%SR
HMMT 202585.8%SR
Beyond AIME69.1%SR
Humanity's Last Exam11.2%SR

AA Evaluation Indices

Intelligence Index
11.9
Gpqa
0.7
Tau2
0.5
Ifbench
0.3
Scicode
0.3
Hle
0.1
Terminalbench Hard
0.0
Lcr
0.0

LLM Stats Category Scores

Language
90
Legal
90
Finance
90
Healthcare
90
Instruction Following
80
Math
80
Physics
80
Structured Output
80
General
80
Biology
80
Chemistry
80
Reasoning
70
Creativity
70
Writing
70
Search
50
Frontend Development
50
Agents
50
Code
50
Vision
10

Pricing

Input Price$0.042 / 1M tokens
Output Price$0.17 / 1M tokens
Blended Price (3:1)$0.074 / 1M tokens

Speed

Tokens/sec117.0
Time to First Token1.24s
Time to Answer18.33s

Provider Price Ranking

Provider Price Ranking

3 providers

Cheapest: FastRouterMost Expensive: NanoGPT
ProviderInputOutput
1FastRouterCheapest
$0.04
$0.16
2SarvamPRIMARY
$0.042
$0.17
3NanoGPT
$0.045
$0.177

Compare pricing across different API providers for this model.

External Sources

LLM StatsArtificial Analysis