Nvidia released Nemotron 3.5 Lightning, a 30-billion parameter open source model designed for specialized tasks within enterprise AI agent systems.

The mixture-of-experts model targets specific functions including code review, tool use, security monitoring, and billing queries. Nvidia released the model weights, training code, and recipes under the Linux Foundation's OpenMDW 1.1 license on August 11.

Nemotron 3.5 Lightning sits between Nvidia's larger Nemotron 3 Ultra (550 billion parameters) and smaller Nemotron 3 Nano (8 billion parameters) within the Nemotron 3 family launched last December.

The model runs on local enterprise hardware including Nvidia's RTX PCs, DGX Spark, and Jetson systems. This addresses growing enterprise concerns about token-based API costs for cloud-hosted models.

Nvidia also shipped NeMo Switchyard, an open source routing library that directs prompts to the most suitable model across an enterprise's mix of open, proprietary, or Nvidia models.

Competing with Chinese AI Labs

The release follows intensified competition from Chinese AI vendors, particularly after Moonshot AI and Alibaba launched powerful open-weight models last month. Analysts consider Alibaba's Qwen 3.8 Max a benchmark for local agentic development.

"It's a testament to just how impactful the Chinese models have been and will be," said Bradley Shimmin, analyst at Futurum Group.

Nvidia CEO Jensen Huang recently defended open source AI models as the Trump administration considers restrictions on Chinese open-weight models, arguing they enable sovereignty and strengthen cybersecurity.

"Models help Nvidia sell its infrastructure better," said Gartner analyst Arun Chandrasekaran. "When you sell a model, you have to run the model somewhere."

The model's local deployment capability addresses enterprise cost concerns around token-based pricing. "Tokenomics over time have already proven to be a little bit hazardous for companies," Shimmin noted, citing lack of transparency in token spending.

Nvidia plans to expand enterprise adoption through its hardware ecosystem as competition in the open source AI model space accelerates.