Sarvam AI Opens Sarvam Token Factory to All Developers
Sarvam AI's Token Factory runs its own and open-weight AI models on GPU clusters in India through one API.
Key Takeaways
- Sarvam AI opened Sarvam Token Factory to all developers and enterprises on November 16, 2026.
- It runs Sarvam 105B and open-weight models such as DeepSeek V4.1 Flash, Kimi K3.1 Pro and Gemma 4 31B on GPUs in India.
- DeepSeek V4.1 Flash Standard has an introductory price of Rs 14 per million input tokens and Rs 55 per million output tokens.
- The DeepSeek V4.1 Flash Interactive endpoint, built for coding agents, costs Rs 25 per million input tokens and Rs 105 per million output tokens.
- Sarvam says prompts stay on GPUs in India.
Table of Contents 7 sections
Sarvam AI launched Sarvam Token Factory for everyone on November 16, 2026, opening its India-hosted AI inference platform to all developers and enterprises. The service runs Sarvam's own models and open-weight models from other AI labs on GPU clusters in India, through one OpenAI-compatible API.
The models on offer include Sarvam 105B, GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, Kimi K3.1 Pro and Google's Gemma 4 31B. Customers can choose dedicated GPU capacity.
Sarvam Token Factory at a glance
| Detail | Value |
|---|---|
| Company | Sarvam AI, Bengaluru |
| Product | Sarvam Token Factory, an AI inference service |
| Open to everyone from | November 16, 2026 |
| Where models run | GPU clusters in India |
| How developers connect | One OpenAI-compatible API |
| Models include | Sarvam 105B, GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, Kimi K3.1 Pro, Gemma 4 31B |
| GPU capacity | Dedicated |
| DeepSeek V4.1 Flash Standard, introductory price | Rs 14 per million input tokens, Rs 55 per million output tokens |
Sarvam Token Factory keeps prompts on GPUs in India
Sarvam says prompts sent to Token Factory run on GPUs in India and are not passed on to the overseas APIs of the companies that built the models. It also runs the same inference stack on premises for what it calls critical institutions, and offers that option to enterprises.
How Sarvam AI got to Token Factory
| When | What happened |
|---|---|
| 2023 | Vivek Raghavan and Pratyush Kumar founded Sarvam AI in Bengaluru |
| November 2026 | Sashikumar Sreedharan joins Sarvam as President and Chief Business Officer |
| November 16, 2026 | Sarvam launches Sarvam Token Factory for everyone |
Sarvam's prices for DeepSeek V4.1 Flash on Token Factory
DeepSeek V4.1 Flash comes through two endpoints. Sarvam says the Interactive endpoint, built for coding agents, runs at about 200 tokens per second per user and costs Rs 25 per million input tokens and Rs 105 per million output tokens. The Standard endpoint, for background and batch work, has an introductory price of Rs 14 for input and Rs 55 for output per million tokens.
Sarvam on building an AI stack without lock-in
"Today, we are launching Sarvam Token Factory for everyone."
"Teams can choose the models that fit their workloads and build an AI stack they own, without lock-in."
Where Sarvam Token Factory stands after launch
Sarvam Token Factory is now open to all developers and enterprises, with Sarvam 105B and open-weight models from DeepSeek, Kimi, GLM and Gemma served from GPUs in India. Customers can choose dedicated GPU capacity.
More news
Frequently Asked Questions
What is Sarvam Token Factory?
When did Sarvam Token Factory open to everyone?
Which models can developers use on Sarvam Token Factory?
How do developers connect to Sarvam Token Factory?
How much does DeepSeek V4.1 Flash cost on Sarvam Token Factory?
Where does Sarvam Token Factory process prompts?
Does Sarvam Token Factory send prompts abroad?
Can enterprises run Sarvam Token Factory's stack on premises?
Who founded Sarvam AI, the company behind Sarvam Token Factory?
How can customers buy GPU capacity on Sarvam Token Factory?
What is Sarvam's Standard endpoint for DeepSeek V4.1 Flash used for?
Who joined Sarvam AI as President in November 2026?
More in AI
SquadStack Launches Maya, a Sales LLM Built and Hosted in India
SquadStack has launched Maya, a large language model it built only for live sales calls.
Fractal Analytics Is in Buyout Talks With Up to 25 AI Startups
Fractal Analytics is in talks with 20 to 25 AI startups about possible acquisitions, its CEO said.
Amazon Launches Alexa for Shopping AI Assistant in India
Amazon has started rolling out Alexa for Shopping, a free AI shopping assistant, to customers in India.
