NVIDIA Releases Its First Open-Source AI Model Nemotron 3.5 Lightning
Nashnova编辑部
Nvidia on August 12 released Nemotron 3.5 Lightning, its first open-source AI model — free for businesses to run and modify on a single GPU. The endgame of free AI is still selling more chips.
What can this model actually do?
Nemotron 3.5 Lightning uses distillation — training a smaller model on a larger model's outputs — to compress stronger capabilities into a package that runs on a single GPU.
Businesses can download, use, and modify it with no license application or fee to Nvidia.
This means → companies can train and deploy AI locally, keeping proprietary data in-house instead of sending it to an AI vendor who might become a competitor.
Why is Nvidia giving a model away for free?
CEO Jensen Huang put it directly: "Free AI should be good for hardware. Free AI should be good for chips."
In plain terms = free model → more enterprises adopt AI → deployments scale up → more GPUs needed to run those models → Nvidia sells more chips.
This reflects a strategic bet: Nvidia's revenue center is not the model itself — it is using open source as a lever to drive hardware volume.
What is NeMo Switchyard?
Nvidia simultaneously released NeMo Switchyard, an open-source routing tool that automatically picks the most suitable — and cheapest — model for each AI-agent task.
VP Kari Briski said Nvidia originally built it to control its own inference costs, and has now opened it externally.
This means → not every task needs a large model. The router continuously balances quality, latency, and cost, helping enterprises push AI spending down.
Where does the open-source camp stand now?
Timing matters: Meta CEO Mark Zuckerberg published an open-source AI manifesto and released coding model Muse Spark just one day earlier. Nvidia followed the next day.
CrowdStrike, CodeRabbit, and Harvey have already completed testing and customized deployments. The model is live on HuggingFace and Nvidia's website.
This reflects a broadening open-source push across Silicon Valley — major players are racing to claim the AI ecosystem's entry points through open source.
How is the market reacting?
ARK Invest bought a combined 122,000 Nvidia shares on August 11. The stock closed down 2.86% at $217.55 that day.
Nvidia's share price recovered 1.6% in pre-market trading on August 12, touching $220.94.
This means → whether open-source models can meaningfully expand end-user GPU demand is the key test of the "free AI sells chips" thesis.
Content is for reference only, not financial advice.