Sarvam AI Unveils Trillion-Parameter Model Plans, Challenges ChatGPT and Gemini at 5x Lower Cost

Sarvam AI Announces Ambitious Trillion-Parameter Model

Indian AI startup Sarvam AI has unveiled plans to develop a trillion-parameter model from scratch, marking a bold entry into competition with global AI leaders including OpenAI's ChatGPT, Google's Gemini, and Anthropic's Claude1

1

. Speaking at the company's inaugural Epoch conference in Bengaluru on Thursday, co-founder Pratyush Kumar revealed that the new model will target advanced use cases including coding, cybersecurity, scientific research, and simulation workloads2

2

. The company expects to deliver this trillion-parameter model within six months, building it entirely from the ground up3

3

.

Aggressive Pricing Strategy Targets Market Disruption

Sarvam AI is positioning its offerings at price points significantly below global competitors. The company announced that its Sarvam 105B model will cost $0.80 per one million blended tokens5

5

. Kumar stated that Sarvam's foundational large language models will be priced five to ten times cheaper than models like ChatGPT or Gemini4

4

. The company's existing voice AI systems are already four to five times cheaper than comparable global offerings, including for English speech recognition1

1

. This aggressive pricing strategy aims to make advanced AI capabilities more accessible while building a competitive edge in the market.

Current Performance Demonstrates Market Readiness

The Indian AI startup's existing 100-billion-parameter model has already demonstrated strong performance metrics. The platform has processed over 325 million minutes of voice calls, and in benchmark testing, performed within four to five percentage points of significantly larger AI models1

1

. The current model focuses on voice AI, conversational systems, English speech recognition, dictation, and simulation-based tasks. Sarvam AI's strongest commercial traction comes from the banking and financial services sector, where institutions are deploying the company's AI stack for voice-based and agent-driven workflows, including an AI agent used by SBI Life's sales network1

1

.

Building End-to-End AI Stack for India

At the Epoch conference, Sarvam AI unveiled a comprehensive end-to-end AI stack spanning foundational models, software services, and hardware. New product launches included an India-hosted inference platform, AI-powered smart glasses under its wearables brand Kaze, a Python SDK for model training, the Bulbul V4 multilingual text-to-speech model, Sarvam Vision 2.0 vision-language model, and Indus, an agentic AI platform integrating work, voice and coding capabilities1

1

. These offerings emphasize Indian language capabilities, positioning the company to serve local markets more effectively than global competitors4

4

.

AI Sovereignty Push Gains Momentum

The announcement comes amid growing concerns about AI sovereignty following President Donald Trump's move to ban exports of latest AI models from Anthropic and OpenAI. Just days before the conference, Kumar had criticized the White House for blocking access to Claude Mythos and Fable5 AI models, stating that countries cannot afford to mistake access to cutting-edge AI systems for true technological ownership3

3

. Kumar emphasized that relying on overseas AI providers could result in dependence not only in terms of costs but also data, adding that "sovereignty should not be a tax"1

1

. Co-founder Vivek Raghavan stated the company aims to build a "real token factory in India," arguing that serving AI inference from within the country represents an important step toward technological self-reliance1

1

.

Strategic Expansion and Talent Acquisition

Sarvam AI is establishing a US office to recruit Indian-origin AI researchers working overseas, part of its strategy to compete globally while maintaining Indian roots1

1

. The company has appointed AI researcher Devendra Singh Chaplot, a founding member of Mistral AI and Thinking Machines Lab, as a part-time advisor based in California5

5

. The startup recently closed a $300 million Series-B funding round led by HCLTech, with participation from Bessemer Venture Partners, Khosla Ventures, and Peak XV Partners, bringing its valuation to $1.5 billion3

3

. Additionally, Sarvam AI announced partnerships with three IITs and two state governments for pilot projects in education and public service delivery5

5

. Kumar stated that Indian AI products should compete globally rather than seek acceptance solely because they are domestically developed, signaling the company's ambition to manufacture intelligence rather than merely consume AI1

1

.