Nvidia Accelerates Open AI Frontier with Broad Model Expansion
March 19, 2026, 10:05 am
Location: United Kingdom, England, South Cambridgeshire
Employees: 501-1000
Founded date: 1994
Nvidia pushes open AI innovation. New models empower AI agents, physical systems, and healthcare. The company expands Nemotron for multimodal intelligence. It introduces Cosmos, GR00T, and Alpamayo for robotics and autonomous vehicles. BioNeMo advances drug discovery and biomedical research. Nvidia forms a coalition for frontier model development. These initiatives underscore a commitment to global AI advancement.
Nvidia unveils a major expansion of its open artificial intelligence portfolio. The company positions open models as critical for the AI economy. It introduces new models across diverse sectors. These advancements aim to accelerate agentic, physical, and healthcare AI development. Nvidia's announcements, made at its Global Technology Conference (GTC), highlight a strategy of global collaboration.
Open models drive innovation. They offer significant cost savings in production. Nvidia is a leading contributor to open-source AI. Its expanded ecosystem provides powerful tools. Developers gain access to cutting-edge AI systems.
The Nemotron family leads agentic AI development. These models support complex AI agents. They move beyond simple question-answering. The new Nemotron 3 lineup integrates language, vision, and voice. These multimodal capabilities enable natural conversations. They facilitate complex reasoning. They extract insights from various data forms.
Nemotron 3 Ultra offers frontier-level performance. It uses Nvidia's NVFP4 floating-point format. This achieves five times greater throughput efficiency. It powers AI-native applications. These include coding assistants and enterprise search. Automated workflows also benefit. Nemotron 3 Ultra is pretrained in NVFP4. It offers superior intelligence. Its max throughput doubles top open base models.
Other Nemotron 3 models include Omni and VoiceChat. Omni combines audio, vision, and language understanding. AI systems interpret documents and video simultaneously. VoiceChat enables real-time conversational interaction. It merges speech recognition, language processing, and text-to-speech.
Reliability is paramount. Nvidia introduces Nemotron safety models. An agentic retrieval pipeline detects unsafe content. It improves response accuracy in multimodal environments.
A growing ecosystem of enterprise software companies adopts Nemotron. Automation Anywhere, CodeRabbit, and CrowdStrike deploy these models. Perplexity AI and ServiceNow also integrate them. LangChain has added Nemotron models to its agent development framework. Edison Scientific uses Nemotron in its Kosmos AI scientist. Kosmos performs hundreds of research tasks daily.
Nvidia supports sovereign AI model development. Developers use Nemotron models to build local AI. These systems serve billions in native languages. They align with local cultures and values. Countries like Singapore, France, India, and Japan benefit. Nvidia also released Nemotron-Personas. These are privacy-preserving synthetic datasets. They are grounded in local census data.
Physical AI sees significant expansion. Nvidia introduces foundation models for robotics and autonomous systems. These models help machines perceive, reason, and act. Cosmos 3 simulates complex environments. It supports physical reasoning for robots. It also aids autonomous machines.
Isaac GR00T N1.7 is a vision-language-action model. It targets humanoid robots. Alpamayo 1.5 improves reasoning and navigation. It enhances autonomous vehicles. HCL Technologies, Johnson & Johnson MedTech, and the Toyota Research Institute use Cosmos. LG Electronics, Neur Robotics, and Noble Machines adopt GR00T. They scale humanoid robotics deployments.
Nvidia CEO Jensen Huang previewed GR00T N2. This next-generation robot foundation model is based on DreamZero research. It uses a new world action model architecture. GR00T N2 helps robots succeed in new tasks. It performs in new environments. Its success rate doubles leading VLA models. GR00T N2 currently ranks highly in robot policy benchmarks. Its release is anticipated by year-end.
Healthcare and life sciences also advance. The BioNeMo platform provides AI tools. It models biological systems. A new generative model, Proteina-Complexa, accelerates drug discovery. It generates protein binders. These attach to target molecules. Novo Nordisk, Viva Biotech, and Manifold Bio use Proteina-Complexa. They design and test generated protein binders.
Nvidia collaborates on critical initiatives. It partnered with Google DeepMind. The European Molecular Biology Laboratory’s European Bioinformatics Institute also joined. Seoul National University contributed. Together, they expanded the AlphaFold protein structure database. Tens of millions of predicted protein complex structures were added. This data helps identify new drug targets. It deepens understanding of disease biology.
Nvidia also introduced nvQSP. This GPU-accelerated simulation engine targets pharmaceutical research. It explores treatment scenarios in computer models. This happens before clinical trials. nvQSP delivers up to 77 times faster performance. It surpasses traditional CPU simulations. Scientists analyze hundreds of dose levels and patient subpopulations. This previously took much longer.
The Nemotron Coalition aims to advance open frontier models. This partnership shares expertise, datasets, and computing infrastructure. Building frontier models demands significant investment. Most organizations cannot manage this alone. The coalition addresses this challenge.
Initial members include Black Forest Labs, AnySphere, and LangChain. Mistral AI, Perplexity, and Reflection AI also participate. The coalition’s first project is a new open base model. Nvidia and Mistral AI co-develop it. It trains on Nvidia’s DGX Cloud infrastructure. This model will be released to the open ecosystem. It will form the foundation for the upcoming Nemotron 4 family. Nvidia's expansive open AI strategy drives innovation across industries. It promises a new era of intelligent systems.
Nvidia unveils a major expansion of its open artificial intelligence portfolio. The company positions open models as critical for the AI economy. It introduces new models across diverse sectors. These advancements aim to accelerate agentic, physical, and healthcare AI development. Nvidia's announcements, made at its Global Technology Conference (GTC), highlight a strategy of global collaboration.
Open models drive innovation. They offer significant cost savings in production. Nvidia is a leading contributor to open-source AI. Its expanded ecosystem provides powerful tools. Developers gain access to cutting-edge AI systems.
The Nemotron family leads agentic AI development. These models support complex AI agents. They move beyond simple question-answering. The new Nemotron 3 lineup integrates language, vision, and voice. These multimodal capabilities enable natural conversations. They facilitate complex reasoning. They extract insights from various data forms.
Nemotron 3 Ultra offers frontier-level performance. It uses Nvidia's NVFP4 floating-point format. This achieves five times greater throughput efficiency. It powers AI-native applications. These include coding assistants and enterprise search. Automated workflows also benefit. Nemotron 3 Ultra is pretrained in NVFP4. It offers superior intelligence. Its max throughput doubles top open base models.
Other Nemotron 3 models include Omni and VoiceChat. Omni combines audio, vision, and language understanding. AI systems interpret documents and video simultaneously. VoiceChat enables real-time conversational interaction. It merges speech recognition, language processing, and text-to-speech.
Reliability is paramount. Nvidia introduces Nemotron safety models. An agentic retrieval pipeline detects unsafe content. It improves response accuracy in multimodal environments.
A growing ecosystem of enterprise software companies adopts Nemotron. Automation Anywhere, CodeRabbit, and CrowdStrike deploy these models. Perplexity AI and ServiceNow also integrate them. LangChain has added Nemotron models to its agent development framework. Edison Scientific uses Nemotron in its Kosmos AI scientist. Kosmos performs hundreds of research tasks daily.
Nvidia supports sovereign AI model development. Developers use Nemotron models to build local AI. These systems serve billions in native languages. They align with local cultures and values. Countries like Singapore, France, India, and Japan benefit. Nvidia also released Nemotron-Personas. These are privacy-preserving synthetic datasets. They are grounded in local census data.
Physical AI sees significant expansion. Nvidia introduces foundation models for robotics and autonomous systems. These models help machines perceive, reason, and act. Cosmos 3 simulates complex environments. It supports physical reasoning for robots. It also aids autonomous machines.
Isaac GR00T N1.7 is a vision-language-action model. It targets humanoid robots. Alpamayo 1.5 improves reasoning and navigation. It enhances autonomous vehicles. HCL Technologies, Johnson & Johnson MedTech, and the Toyota Research Institute use Cosmos. LG Electronics, Neur Robotics, and Noble Machines adopt GR00T. They scale humanoid robotics deployments.
Nvidia CEO Jensen Huang previewed GR00T N2. This next-generation robot foundation model is based on DreamZero research. It uses a new world action model architecture. GR00T N2 helps robots succeed in new tasks. It performs in new environments. Its success rate doubles leading VLA models. GR00T N2 currently ranks highly in robot policy benchmarks. Its release is anticipated by year-end.
Healthcare and life sciences also advance. The BioNeMo platform provides AI tools. It models biological systems. A new generative model, Proteina-Complexa, accelerates drug discovery. It generates protein binders. These attach to target molecules. Novo Nordisk, Viva Biotech, and Manifold Bio use Proteina-Complexa. They design and test generated protein binders.
Nvidia collaborates on critical initiatives. It partnered with Google DeepMind. The European Molecular Biology Laboratory’s European Bioinformatics Institute also joined. Seoul National University contributed. Together, they expanded the AlphaFold protein structure database. Tens of millions of predicted protein complex structures were added. This data helps identify new drug targets. It deepens understanding of disease biology.
Nvidia also introduced nvQSP. This GPU-accelerated simulation engine targets pharmaceutical research. It explores treatment scenarios in computer models. This happens before clinical trials. nvQSP delivers up to 77 times faster performance. It surpasses traditional CPU simulations. Scientists analyze hundreds of dose levels and patient subpopulations. This previously took much longer.
The Nemotron Coalition aims to advance open frontier models. This partnership shares expertise, datasets, and computing infrastructure. Building frontier models demands significant investment. Most organizations cannot manage this alone. The coalition addresses this challenge.
Initial members include Black Forest Labs, AnySphere, and LangChain. Mistral AI, Perplexity, and Reflection AI also participate. The coalition’s first project is a new open base model. Nvidia and Mistral AI co-develop it. It trains on Nvidia’s DGX Cloud infrastructure. This model will be released to the open ecosystem. It will form the foundation for the upcoming Nemotron 4 family. Nvidia's expansive open AI strategy drives innovation across industries. It promises a new era of intelligent systems.