Nvidia ropes in Sarvam, Mistral, Perplexity in global ‘Nemotron Coalition’ to build open AI models

In a significant stride towards fostering open and sovereign artificial intelligence, Nvidia has formed a collaborative group known as the Nemotron Coalition, which includes several prominent global AI laboratories, such as India's Sarvam. This initiative was unveiled during Nvidia's GTC conference and aims to develop next-generation open foundation models, positioning Sarvam alongside esteemed entities like Mistral AI, Perplexity, Thinking Machines Lab, LangChain, Cursor, Reflection AI, and Black Forest Labs. Sarvam is set to leverage its expertise in creating multilingual and voice-first AI models tailored to local contexts, filling a crucial gap in the current AI landscape. Pratyush Kumar, co-founder and CEO of Sarvam, emphasized the importance of inclusivity in AI, stating, "AI reaches its full potential when it works in every language and for every community." He highlighted that open models empower developers to tailor advanced capabilities to meet real-world demands effectively. Sarvam's participation indicates India's expanding influence in the global AI dialogue, particularly in areas such as multilingual AI and voice interfaces, essential for reaching larger populations. This coalition's commitment to open models aligns perfectly with India’s own initiatives, like the IndiaAI Mission, which advocates for the development of foundational models independent of foreign systems. The Nemotron Coalition is described as a groundbreaking partnership aimed at pooling resources such as data, computational power, evaluation frameworks, and research expertise to create advanced models within an open ecosystem. Nvidia's CEO, Jensen Huang, remarked, "Open models are the lifeblood of innovation and the engine of global participation in the AI revolution," underscoring the coalition's mission to foster transparency, collaboration, and sovereignty in AI development. The first model to emerge from this collaboration will be co-created by Nvidia and French AI firm Mistral AI, utilizing Nvidia’s DGX Cloud infrastructure. This initial model will serve as the foundation for the future Nemotron 4 family of open models. This move comes at a time when the global discourse on open versus proprietary AI systems is intensifying, with leading companies like OpenAI and Google largely keeping their advanced models under wraps. In contrast, Nvidia's coalition seeks to establish a shared base model that can be customized by developers and enterprises according to specific industries and use cases. Arthur Mensch, CEO of Mistral AI, stated that open frontier models are essential for AI to serve as a true platform. Additionally, Perplexity's CEO, Aravind Srinivas, noted that open models offer developers the flexibility to enhance performance and drive AI applications into everyday scenarios. The coalition is also focused on advancing practical applications of AI, particularly in developing agentic AI systems. Harrison Chase, CEO of LangChain, mentioned that with the significant adoption of LangChain’s frameworks, it is crucial for frontier models to go beyond mere intelligence to enable reliable tool usage and effective agent coordination. Each member of the coalition is expected to bring unique capabilities to the table, enhancing the collective effort to revolutionize AI development.