AI & Innovation

Sarvam Token Factory launched: eight AI models hosted in India, rupee pricing

By Anika SharmaAI & Innovations Editor, Peepals Tech
WhatsAppX
Representative image: laptop keyboard

Representative image: Pexels / Pixabay

Sarvam AI launched Token Factory on 16 November 2026, offering eight AI models on servers in India through one OpenAI-compatible API, priced in rupees.

Sarvam Token Factory: what happened

Sarvam Token Factory, a new service from Indian AI company Sarvam AI, went live for everyone on Monday, 16 November 2026. It is an inference platform: a service that runs ready-made AI models on its own computers and lets developers send them requests over the internet, paying for what they use. Token Factory offers eight models through a single application programming interface (API) that works the same way as OpenAI's, and all of them run on graphics-processor (GPU) servers in India. In its launch post, Sarvam says it matches or beats leading global inference providers on speed and cost for most models, a claim that has not been independently tested.

Sarvam Token Factory: key facts

  • Launch: Monday, 16 November 2026; generally available to all users.
  • What it is: a pay-per-use inference service with one OpenAI-compatible API for eight models.
  • Where it runs: GPU infrastructure located in India.
  • Catalogue at launch, as listed by one industry newsletter: Sarvam 105B, GLM 5.3, GLM 5.3 Flash, DeepSeek V4.1 Flash, DeepSeek V4 Pro, Kimi K3, Kimi K3.1 Pro and Gemma 4 31B.
  • Pricing: in rupees; Sarvam 105B is listed at Rs 10 and Rs 25 per million tokens (a token is a small piece of text, roughly part of a word).
  • Data: reports say the service keeps no customer data after a request is answered (zero data retention).
  • Criticism: some in the industry call it a hosting business that mainly resells foreign open-weight models, many of them Chinese.

Sarvam Token Factory: why it matters

Most Indian companies that use AI today send their data to servers abroad, run by US cloud and AI firms. For banks, hospitals and government bodies, that raises questions about where sensitive data goes and which country's laws apply. A service that runs both Indian and popular open models on servers inside India, with prices in rupees and no data kept after each request, answers part of that concern. It also makes switching easy: because the API copies OpenAI's format, a developer can often move an existing app by changing an address and a key.

The launch has reopened a long-running argument about what "sovereign AI" should mean in India. Sarvam is known as a builder of Indian models: it launched its own 30-billion and 105-billion-parameter models earlier this year. Token Factory, however, puts its flagship next to models developed abroad, and six of the eight listed models come from the GLM, DeepSeek and Kimi families. Supporters say that giving Indian users fast, local access to the best open models is itself a sovereign capability. Critics say it makes Sarvam look more like a reseller than a model-maker.

For developers, the practical question is cost and speed. Rupee pricing removes currency risk for Indian start-ups, and local hosting can reduce delay for users in India. Whether Token Factory is cheaper than overseas rivals will depend on each model and workload, and Sarvam's comparison is its own.

Peepals calculation: six of the eight launch models come from the GLM, DeepSeek and Kimi families, two each, which is 75% of the catalogue (6 ÷ 8 = 0.75). And Sarvam 105B's higher listed price is 2.5 times its lower one (25 ÷ 10). Our calculation from the reported figures.

ItemDetail
Launch date16 November 2026
Models at launch8
API styleOpenAI-compatible
HostingGPU servers in India
Sarvam 105B price (listed)Rs 10 and Rs 25 per million tokens

Sarvam Token Factory: impact

Short-term: Indian developers and companies can try open models hosted in India with rupee billing, and compare speed and cost against overseas providers.

Long-term: If enterprises and government users adopt it, Token Factory could become a base layer for Indian AI apps, and push the debate on whether sovereign AI means building models or controlling where they run.

Who is affected: Indian start-ups and enterprises building AI apps, regulated sectors such as banking and health that care where data is processed, rival cloud and AI providers, and the IndiaAI programme.

Sarvam Token Factory: key takeaway

If your company needs AI with data kept in India, Token Factory is worth testing, but check Sarvam's speed and cost claims against your own workload before switching.

Sarvam Token Factory: questions and answers

What is Sarvam Token Factory?

It is a pay-per-use service from Sarvam AI, launched on 16 November 2026, that runs eight AI models on servers in India and lets developers reach them through one OpenAI-compatible API.

Which models are on Sarvam Token Factory?

The launch list includes Sarvam 105B along with open-weight models from the GLM, DeepSeek, Kimi and Gemma families, according to an industry newsletter's listing.

How much does it cost?

Prices are in rupees; Sarvam 105B is listed at Rs 10 and Rs 25 per million tokens, and full per-model prices are on Sarvam's pricing page.

Why are some people critical of it?

Critics say most of the models are developed abroad, many in China, so the service looks more like hosting and reselling than building India's own AI.

Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.

#Sarvam Token Factory#Sarvam AI#sovereign AI India#AI inference India#Sarvam 105B

By Anika Sharma, AI & Innovations EditorCovers AI & Innovation and more for Peepals Tech. All stories by Anika Sharma

Your week ahead, every Sunday. Free.

Exam and result dates, IPOs opening, scheme deadlines, government holidays and the RBI decision for the next 7 days, plus the week's top stories. One email a week, unsubscribe in one click.

Anika Sharma is a pen name of the Peepals Global Editorial Team, which researches, fact-checks and publishes every story. The portrait is an illustration.

Responsible for this story: the Peepals Global Editorial Team, led by editor Prashant Sonar. How we report

More from Peepals Tech

Your week ahead, every Sunday. Free.

Exam and result dates, IPOs opening, scheme deadlines, government holidays and the RBI decision for the next 7 days, plus the week's top stories. One email a week, unsubscribe in one click.