Despite Indian AI startup Sarvam being on the move to expand its operations with its latest flagship unveiling of Large Language Model (LLM), Sarvam-M, it’s not steered clear of the backlash coming its way. And so, Zoho CEO Sridhar Vembu came to the company’s rescue this weekend.

As reviews for Sarvam AI‘s Sarvam-M, a 24-billion-parameter open-weights hybrid language model built on top of Mistral Small, continued to pour in on social media, the verdict was that international techies weren’t fully on board. Despite the company’s India-focussed game plan, comparisons with OpenAI or DeepSeek sought to question its position on the global mainstream.

Sarvam-M backlash on X

The company was proud of its new development, with co-founder Vivek Raghavan saying, “Sarvam-M represents an important stepping stone on our journey to build Sovereign AI for India.” However, Deedy Das, an investor at Menlo Ventures, prompted a heated debate after blasting the advancement as “embarrassing,” drawing stark parallels with an open-source model, which was developed by two Korean college students.

ALSO READ

“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,” he tweeted early Saturday (IST). “In contrast. 2 Korean college trained an open-source model that did ~200k last month.”

As per his observations, the Indian AI scene seems to be more about doing “cool AI thing that cool AI people do,” and not “let’s solve important hard problems.” He added in a follow-up tweet: “No one is asking for a slightly better 24B indic model. Clearly.”

Concluding his review-bombing thread on X, he finally said that his criticism of the model had been misconstrued, as he wasn’t essentially “against trying… against Sarvam” or even India building AI. “These are ridiculous things to insinuate,” he defended his stance, adding, “I’m disappointed at their direction and expected them to accomplish more with their resources.”

Zoho CEO defends Sarvam AI, others join

Zoho’s Sridhar Vembu flipped the dialogue, noting that instant virality was impossible, and not necessarily the top priority, in this sphere. “In defence of Sarvam.ai, I will point out that there is no product we have built that was ever an instant hit.”

Citing his own side’s example, he went on to say in an X post, “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.” Encouraging the Sarvam team to “keep fighting the good fight,” he argued that instant success was “neither necessary nor sufficient to succeed long term.”

ALSO READ

Sarvam AI’s own Harveen Singh Chadha noted that while many people were opting to launch diatribes against the model on their social media platforms, a significant chunk of the faction didn’t even attempt to follow up with feedback after trying the model on their own. “People with 90k following will post a lot of gyaan on twitter on a Saturday morning. But won’t take out 10 minutes to try the model and share feedback. This is the reason why not enough people care, if you just use your following to criticise without even trying you are a part of problem.”

Aashay Sachdeva, also from Sarvam AI, also backed the model, calling it “not just a good indic model,” but also a “good overall 24B model with thinking mode.” Directly snapping back at Deedy Das’ critique, he also wrote, “And yes we want to do cool things in ML for the sake of it.”

Sarvam AI describes its new introduction of Sarvam-M as “a hybrid Indic model fine-tuned for Indian languages, math and programming.” The website’s official blog post details that the model has been enhanced through a three-step process involving, Supervised Fine-Tuning (SFT), Reinforcement Learning with Verifiable Rewards (RLVR) and Inference Optimisations. Sarvam-M is currently available through Sarvam’s API. It can be downloaded from Hugging Face.