Tools & Products

Nvidia Releases Nemotron 3 Open Models for Agentic Workflows

Hybrid architecture combines Mamba and Transformer designs to optimize autonomous system efficiency.

Olivia Sharp 2 min read 735 views
Free
Nvidia launched the Nemotron 3 model family on December 15, utilizing a hybrid architecture to power efficient agentic AI and sovereign deployments.

Nvidia formally entered the open-model software market on December 15, 2025, with the release of the Nemotron 3 family. The launch includes three variants—Nano, Super, and Ultra—designed to provide the foundational logic for "agentic" AI systems. By utilizing a novel hybrid architecture, Nvidia aims to address the specific throughput and latency challenges associated with multi-agent orchestration, where models must plan, execute, and iterate without human intervention.

Hybrid Mamba-Transformer Architecture

The defining technical feature of the Nemotron 3 series is its departure from pure Transformer architectures. The models utilize a hybrid design that integrates Mamba State Space Models (SSMs) with …

Archive Access

This article is older than 24 hours. Create a free account to access our 7-day archive.

Share this article

Related Articles