Sarvam AI launches Token Factory, an India-hosted inference platform for its own and open models
Sarvam AI has launched Token Factory, which gives developers and enterprises access to Sarvam 105B and open models such as DeepSeek, GLM, Kimi and Gemma through one API. Inc42 reports that the models run on Sarvam-operated GPUs in India. Pricing and performance figures have not been disclosed.
Bengaluru's Sarvam AI has launched Token Factory, an inference platform that lets developers and enterprises call multiple AI models through a single API. Customers do not have to build or manage their own GPU infrastructure. The launch was reported by Inc42. [1]
The catalogue includes Sarvam's own Sarvam 105B model alongside open models including DeepSeek, GLM, Kimi and Gemma. According to Inc42, inference runs on GPUs that Sarvam operates in India. [1]
Sarvam says its serving stack is faster and more efficient because it uses cache-aware routing, separates prefill and decode infrastructure, and caches across multiple tiers. The company offers both serverless and dedicated capacity. No independent benchmarks of these claims were available. [1]
Inc42 frames in-country inference as a possible draw for Indian businesses with sensitive workloads, as data residency and compliance carry more weight. The same Inc42 edition reports that OpenAI is introducing a program that lets businesses process supported GPT-6 Luna workloads within India. That move would put a global provider into the locally hosted segment. [1]
Inc42 also lists other competitors. It names OpenRouter, which it describes as recently acquired by Stripe and which routes developers to hundreds of models through a unified API. It also names Indian players JarvisLabs and Simplismart, which it says are building their own inference offerings. [1]
Inc42's analysis is that Sarvam's differentiation rests on combining its own models, India-based infrastructure and an optimised serving stack. Model access is becoming commoditised, and Inc42 says Sarvam still has to show competitive latency, token costs and reliability at enterprise scale. [1]
Sources
- Inc42: AI’s Social Layer & Sarvam’s Token FactorySource published · TechDots found it
Statements labelled as a company's claim have not been independently confirmed by TechDots.
Get Startups stories by email
Related coverage
Humyn Labs, an Indian startup supplying robotics training data, raises $47M with US customers driving revenue
Humyn Labs, founded in February, has raised $47 million to supply training data to U.S. robotics developers, TechCrunch reports. About 90% of its revenue comes from the U.S., and Accel is among its backers.
YourStory's SheSparks 2027 expands to two days in Delhi with 'Multiplier Effect' theme
YourStory says its women-focused SheSparks event returns to The Ashok Hotel in Delhi on February 19–20, 2027, its first two-day edition, with investor matchmaking, a returnship fair and a list recognising 100 women.
Fatehpuria Hygiene raises Rs 100 crore in first equity round led by Madhu Kela
Supply chain and fulfilment firm Fatehpuria Hygiene says it has raised Rs 100 crore in its first equity funding, led by investor Madhu Kela, and plans to expand to more than 50 cities over two years, according to YourStory.