E-Paper
Indian AI models outperform OpenAI, Google on key benchmarks: Ashwini Vaishnaw
Ashwini Vaishnaw said homegrown platforms, which will be showcased at the AI Impact Summit in New Delhi next week, has reinforced confidence.
New Delhi : Indian AI models perform better than their counterparts from firms such as OpenAI, Google and High-Flyer on many parameters, IT minister Ashwini Vaishnaw said on Saturday, adding that the these homegrown platforms, which will be showcased at the AI Impact Summit in New Delhi next week, has reinforced confidence in India’s ability to develop models using limited resources.
“In the AI Summit, we will be launching our bouquet of sovereign models… some of the models which have been tested… on so many parameters the models are better than OpenAI [GPT], [Google’s] Gemini Pro and [High-Flyer’s] Deepseek. And that is giving us the confidence that we can develop models on very frugal, innovative, setups and resources…they can actually beat world models,” Vaishnaw said during the Times Group’s ET NOW Global Business Summit.
During its 14-day AI product shipping, which began earlier this month, Indian startup Sarvam unveiled several models, claiming that some outperform global peers.
The firm said that its Saaras V3 model beat Deepgram Nova-3, ElevenLabs Scribe v2, Gemini 3 Pro and GPT-4o Transcribe across the 10 most spoken Indian languages on the IndicVoices benchmark while maintaining low latency in real time.
Sarvam Vision also posted 84.3% accuracy on the English subset of olmOCR-Bench surpassing Gemini 3 Pro and DeepSeek OCR 2.
Prime Minister Narendra Modi is expected to launch some of the sovereign models at the AI Impact Summit, said the Information Technology minister.
