US EditionAI Chipmaker Nvidia Just Released Open Source Models For AI Agents - NVIDIA (NASDAQ:NVDA)
Nvidia's Nemotron 3 models use mixture of experts technology, offering up to 4x speed improvement and ranked most efficient by Artificial Analysis.
- On 2025-12-16, Nvidia launched the open‑source Nemotron 3 Nano, Super and Ultra to power AI agents, with the Nano now available on HuggingFace.
- Nvidia used the mixture‑of‑experts technique to design Nemotron 3 and has built models since Megatron‑LM in 2019, a strategy Kari Briski said pushes its hardware to the limit.
- The Nemotron 3 Nano provides a 1 million‑token context window and Nvidia said it can be up to 4x more performant than Nemotron 2 Nano with 60% fewer reasoning tokens.
- Developers can run Nemotron models as Nvidia NIM microservices or via Amazon Bedrock and, soon, Google Cloud, Coreweave, Nebius, Nscale and Yotta, while Baseten, Deepinfra, Fireworks, FriendliAI, OpenRouter and Together AI will offer or tune them.
- The Super and Ultra follow in the first half of 2026, with Nemotron 3 Super at 100 billion parameters and Nemotron 3 Ultra at 500 billion parameters , the most expensive to run.
12 Articles
12 Articles
One News Page
PYMNTS.comNvidia Bets on Open Models to Power AI Agents
Nvidia announced on Monday (Dec. 15) the debut of the Nemotron 3 family of open models designed to power transparent, efficient and specialized agent-centric AI across industries. The lineup includes Nano, Super and Ultra models that leverage a hybrid latent mixture-of-experts (MoE)…
Nvidia's Launches the Next Generation of Its Nemotron Models
Nvidia today launched its latest set of AI models in its open source Nemotron series for powering AI agents: Nemotron 3 Nano, Super and Ultra. For the first time, Nvidia is also releasing not just the models but also three trillion tokens worth of pre-training data and 18 million samples of post-training data. Thanks to Nvidia’s existing training environments, for which the company is also launching 10 new gym environments, and the company’s ope…
Nvidia announced the family of models, data and open libraries Nemotron 3 designed to boost the development of Artificial Intelligence (IA) with Agents. The Agentica AI is considered as the next step in the evolution of AI, which promises new capabilities for autonomy and automation of processes and complete tasks. In a statement, Nvidia points out that Nemotron 3 models available in Nano, Super and Ultra sizes, introduce a new hybrid architectu…
Nvidia has introduced a new family of open Nemotron 3 artificial intelligence (AI) models that, composed of Nano, Super and Ultra sizes, use a hybrid expert mix architecture (MoE) that drives the development and implementation of more transparent and efficient multiagent systems.
Coverage Details
Bias Distribution
- 100% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
