Details
- NVIDIA launched open models, massive datasets and tools including Nemotron for agentic AI, Cosmos for physical AI, Alpamayo for autonomous vehicles, Isaac GR00T for robotics and Clara for biomedical applications, announced at CES 2026.
- Key players adopting include Bosch, ServiceNow, Palantir, CrowdStrike, Uber, Salesforce, Franka Robotics and Hitachi; datasets feature 10T language tokens, 500K robotics trajectories, 455K protein structures and 100TB vehicle data.
- New features: Nemotron Speech for 10x faster real-time ASR, RAG for multimodal retrieval, Safety models for PII detection; Cosmos Reason 2 tops leaderboards for robot reasoning; Alpamayo 1 as first open AV reasoning VLA; Clara models like La-Proteina for protein design.
- Builds on prior Nemotron 3 and Cosmos from 2025, expanding to speech/multimodal/safety; contrasts with closed models by providing open-source frameworks, blueprints and NIM microservices on Hugging Face/GitHub.
- Part of CES 2026 alongside Rubin AI platform succeeding Blackwell; leads robotics downloads on Hugging Face, accelerates physical AI validation with synthetic data, reducing development time versus real-world collection.
Impact
NVIDIA's open models democratize physical AI development across industries, topping leaderboards and enabling rapid innovation in robotics, AVs and biomedicine while countering closed ecosystems from OpenAI and Google. Early adopters like Palantir and Uber signal strong enterprise traction, positioning NVIDIA to dominate agentic and world-model AI amid rising competition from custom chips by Microsoft and Amazon. This scales AI from chatbots to real-world deployment, potentially slashing costs and timelines for safe, scalable systems.
