No product was ever an instant hit: Zoho’s Sridhar Vembu backs Sarvam AI amid criticism

Zoho's top boss Sridhar Vembu has defended homegrown AI startup Sarvam AI following backlash over the company’s newly released large language model (LLM) Sarvam-M.

“In defense of Sarvam AI, I will point out that there is no product we have built that was ever an instant hit. Even when we were the first mover in a new market and we had done a lot of technical work, we only got slow traction. Instant success is neither necessary nor sufficient to succeed long term,” said Vembu, Co-founder, Zoho, in a post on X (formerly Twitter). 

“Keeping fighting the good fight Sarvam team,” he added. 

The model in question is Sarvam-M (where M stands for Mistral), a 24-billion-parameter open-weights hybrid LLM based on Mistral Small, an open-source language model developed by the French startup Mistral AI.

Sarvam AI stated in a blog that the new model designed for Indian languages is capable of handling reasoning tasks in mathematics and programming.

Also Read

Sarvam AI brings 24B-parameter LLM for Indian languages, reasoning

The launch of Sarvam-M has evoked mixed reactions online, especially from international developers and investors, with many questioning the model’s originality and competitiveness.

Deedy Das, Partner at Menlo Ventures, drew comparisons with a Korean open-source model developed by two college students. 

“India's biggest AI startup, $1B Sarvam, just launched its flagship LLM. It's a 24B Mistral small post trained on Indic data with a mere 23 downloads 2 days after launch. In contrast, 2 Korean college trained an open-source model that did ~200k last month. Embarrassing,” read Das’s post on X. 

Clarifying his comments, Das later said his critique was aimed at the company’s direction and he'd expected it to accomplish more with its resources. 

“I want to be very clear because criticism seems to be very misconstrued. I’m not against trying. I’m not against Sarvam. I’m not saying India shouldn’t build AI. These are ridiculous things to insinuate.  I’m disappointed at their direction and expected them to accomplish more with their resources,” he said. 

Responding to Das's post, Sarvam AI’s Aashay Sachdeva stated that the model was released just the night before and that Hugging Face metrics often take time to update.

“We released it yesterday night, so not 2 days. Hugging Face numbers take a while to update.... and we are definitely not valued at $1B. Yes, we want to do cool things in ML for the sake of it. It is also not just a good (ndic model, it is always a good overall 24B model with thinking mode,” Sachdeva said.

Hugging Face is a platform that offers tools for working with machine learning models, particularly those used for natural language processing.

In late April, the Indian government chose Sarvam AI to build the country’s sovereign LLM as part of the IndiaAI Mission, a national effort to strengthen domestic capabilities in emerging technologies.