DevOps / SRE / Platform · 11.08.2026, 21:55 UTC
Nvidia launches a smaller, faster Nemotron model and a router to put it to work
| Schweregrad | info |
|---|---|
| Kategorie | DevOps / SRE / Platform |
| Quelle | The New Stack ↗ |
| Veröffentlicht | 11.08.2026 UTC |
Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.
Nvidia on Tuesday launched Nemotron 3.5 Lightning, the newest member of its Nemotron 3 family of open models. In addition, the company also launched NeMo Switchyard, a new open-source library that can power model routers.
Nemotron 3.5 Lightning is a 30-billion-parameter mixture-of-experts model that was developed with contributions from the Nemotron coalition and that has reasoning capabilities that come close to the much larger Nemotron 3 Super model. Both Nvidia models, however, do trail the similarly sized Gemma 4 31B from Google on Artificial Analysis’ Intelligence Index.
Credit: Nvidia.
Built for speed
But the focus here, Nvidia argues, isn’t on the benchmarks so much as on speed and the ability to easily customize the model. As with similar small models, the core use case is for a frontier model to plan and orchestrate the work, and for the smaller — and potentially fine-tuned — model to handle execution.
As Nvidia Senior Director Joey Conway told The New Stack recently, the company believes that it’s these systems of models that are the future of AI, after all.
As for speed, Nvidia argues that 3.5 Lightning can deliver up to 4x faster output speeds but as Nvidia’s Kari Briski noted during a press briefing ahead of the announcement, it’s the ability to modify and optimize the model for specific workflows that Nvidia hopes will set it apart from the competition.
Credit: Nvidia.
Post-training for specialized tasks
“General agentic benchmarks are just the starting point, but in production, what matters is accuracy on your task — and that’s where post-training …