Sarvam AI Building One Trillion Plus Parameter Model, 5.5 Times Cheaper Than Gemini, ChatGPT

Sarvam is retailing its flagship Sarvam 105B model at $0.80 per one million blended tokens, compared to $4.50 for OpenAI's GPT-5.4 Mini.

Advertisement
Read Time: 3 mins
Sarvam 105B performed on par with GLM 5.2, a 753-billion-parameter model.
Photo Source: AI-Generated

Sarvam AI will be building a foundational one trillion plus parameter model that will be 5.5 times cheaper than its leading peers across the globe, according to reports on Thursday.

The AI model is expect to display exceptional performance in the realms of coding, cybersecurity and scientific research workloads, and will be built from scratch in India.

Advertisement

ALSO READ: Samsung's Chip Profit Soared 250-Fold On AI Memory Shortages. Why Are Investors Not Celebrating?

"We are very happy to announce that we are building a trillion-plus parameter model right here in India. We are building them from scratch to be competitive in coding, cybersecurity, simulation, science and more," said Pratyush Kumar, co-founder of Sarvam, at the company's launch event in Bengaluru.

This development comes amid AI firms such as Google competing with one another to produce models that have the lowest inference costs in order to drive down expense for enterprise customers.

The company is retailing its flagship Sarvam 105B model at $0.80 per one million blended tokens, compared to $4.50 for OpenAI's GPT-5.4 Mini and $9 for Gemini 3.5 Flash.

Advertisement

According to the firm, this makes their models 5 .5 times cheaper than GPT-5.4 Mini and over 11 times more affordable than Gemini 3.5 Flash.

As per the company, the model is hosted in India, and has commenced work on serving voice AI workloads, processing up to "crores of minutes per day."

Advertisement

"If you are building a voice product today, Sarvam is the cheapest, fastest and most scalable solution," the firm stated. It also claimed that AI models such as Gemini and ChatGPT did not have capabilities at the same level of scale.

Sarvam said its 105-billion-parameter model delivers performance comparable to much larger AI models at a significantly lower cost.

The firm stated that Sarvam 105B performed on par with GLM 5.2, a 753-billion-parameter model, across enterprise benchmarks such as tool use, robustness, consistency and process execution, despite the wide gap in parameter size.

Alongside its language models, Sarvam announced the commercial launch of Sarvam Vision Edge, a document intelligence platform.

The startup said it has partnered with the Odisha government on a research initiative to digitise land records and extract information using the vision model.

ALSO READ: Who's Devendra Singh Chaplot? IIT Bombay Alumnus Who Worked With Musk's xAI Joins Sarvam As Adviser

Sarvam also launched Sarvam Vision 2.0, which it said enhances optical character recognition (OCR) capabilities for enterprise document processing.

Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.


Loading...