Company models

• Earth 2 Medium Range (Atlas architecture) Gen 4

  Earth-2 Medium Range is NVIDIA’s Atlas-based global forecast model that predicts over 70 weather variables up to 15 days ahead, outperforming leading open models like GenCast while running efficiently on GPUs.
  NewImage

  Released 3d ago
• Earth 2 Nowcasting Gen 4

  Earth-2 Nowcasting is NVIDIA’s generative AI storm model that turns satellite and radar data into zero- to six-hour local forecasts in minutes, producing high-fidelity hazardous weather predictions at kilometer scale
  NewVideo

  Released 5d ago
• PersonaPlex 7B v1 Gen 3

  PersonaPlex is NVIDIA’s full duplex conversational AI that lets you pick any role and voice, then maintains that persona through natural, low latency dialogue with interruptions, backchannels and realistic conversational rhythm.
  NewMultimodal

  Released 10d ago
• Nemotron 3 Gen 7
  NewText

  Released 1mo ago
• Nemotron 3 Nano Gen 7
  NewText

  Released 1mo ago
• Orchestrator-8B Gen 7

  ToolOrchestra is NVIDIA’s method for training small “orchestrator” models that decide when and how to call tools and other LLMs. The main model, Orchestrator-8B, beats GPT-5 on hard agent benchmarks like Humanity’s Last Exam while using about 30% of the cost and running roughly 2.5× faster.
  NewText

  Released 2mo ago
• ChronoEdit 14B Diffusers Upscaler LoRA Gen 4

  ChronoEdit-14B is a diffusion-based image and video editor with temporal consistency and built-in upscaling. It supports masked edits, 2x-4x enhancement, and LoRA style adapters for fast, repeatable looks.
  Image

  Released 3mo ago
• Alpamayo R1 Gen 7

  Alpamayo-R1 is NVIDIA's vision-language-action model for autonomous driving, combining Chain of Causation reasoning with trajectory planning to handle long-tail, safety-critical situations more robustly and with interpretable decisions.
  Text

  Released 3mo ago
• Cosmos Predict Gen 4

  Cosmos Predict is NVIDIA’s generative world-model for forecasting what happens next. From short video or sensor histories (plus optional actions), it predicts future frames, object trajectories, and scene state—with uncertainty estimates—so robots, AV stacks, and vision agents can plan ahead.
  Video

  Released 4mo ago
• Nemotron-4 Gen 7

  Nemotron-4 is NVIDIA’s open-weight LLM family (from compact to 340B) built for high-quality reasoning, coding, and synthetic data generation. It supports tool/function calling, JSON outputs, and long-context variants, and is production-ready via NVIDIA NIM/TensorRT-LLM for fast, scalable deployment.
  Text

  Released 4mo ago
• Cosmos Reason Gen 3

  Cosmos Reason is NVIDIA’s open, customizable 7B-parameter reasoning VLM for physical AI and robotics. It learns space/time/physics “common sense,” plans actions from video/images, and was trained with SFT + RL. Available under the NVIDIA Open Model License on GitHub/Hugging Face and as an NVIDIA NIM.
  Text

  Released 4mo ago
• Nemotron Nano 2 Gen 7

  Nemotron-Nano 2 is NVIDIA’s ultra-compact LLM tuned for on-device and edge deployment. It delivers fast instruction following, coding help, and reasoning with low memory use, supports tool/function calling and structured (JSON) outputs, and runs efficiently on Jetson, RTX laptops/desktops, and server GPUs.
  Text

  Released 5mo ago