Sarvam AI

Sarvam-105B

Open WeightsMoEapache-2.0Released Jun 2026

Sarvam AI's sovereign 105B-parameter Mixture-of-Experts model activating ~9B parameters per token, with a 128K-token context window. Trained on 12 trillion tokens across 22 Indian languages using 128 sparse experts with Multi-head Latent Attention and a custom low-fertility Indic tokenizer. Wins the majority of pairwise comparisons on Indian-language and STEM benchmarks.

Capabilities

chatcompletionfunction-callingcode-generationreasoning

Modalities

textcode

Context Window

128K tokens

Providers

available

Available from 1 provider

Cheapest

Sarvam AI

$4.00/1M tokens

Sarvam AI

Providers (1)

Sorted by total cost (input + output per 1M tokens). Click a row to view provider details.

Provider	Pricing (per 1M)	Rate Limits	Regions	Health	Latency
Sarvam AI	In: $1.00Out: $3.00	60 RPM / 120K TPM	ap-south-1	Healthy	0ms

Quick Start

Use this model via Sarvam AI with an OpenAI-compatible SDK.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.sarvam.com/v1",
  apiKey: process.env.SARVAM_API_KEY,
});

const response = await client.chat.completions.create({
  model: "sarvam-105b",
  messages: [
    { role: "user", content: "Hello!" }
  ],
});

console.log(response.choices[0].message.content);

Using Sarvam AI API • OpenAI-compatible SDK