Glean has added support for NVIDIA Nemotron 3 Ultra to its enterprise AI platform, expanding the range of models available to customers.
The new option is aimed at agentic work in businesses, giving customers access to an open model alongside existing proprietary and open-source choices. The platform now provides access to more than 30 models.
The announcement reflects a broader debate in corporate AI over whether companies should standardise on one large language model provider or use several models for different tasks. Glean is aligning itself with the latter approach, arguing that businesses want to choose models based on workload, price, and performance rather than route every job through a single family of systems.
Open models have gained ground as companies look for lower-cost ways to deploy generative AI in routine internal work. Cost control has become more important as usage rises, particularly in applications involving repeated search, drafting, and workflow automation across large employee bases.
According to Glean, Nemotron 3 Ultra offers strong agentic performance for everyday enterprise use. The company said the model delivers 91% of frontier large language model completeness while retaining the cost profile of an open model.
Glean also linked the launch to its existing work with NVIDIA's Nemotron family. Its Waldo agentic search model is post-trained on NVIDIA Nemotron 3 Nano, and the company said Waldo delivers 50% lower latency while using 25% fewer tokens.
Model choice
For Glean, the addition fits a model-agnostic strategy that has become central to how many enterprise software suppliers are approaching AI. Rather than building around a single supplier, vendors are increasingly giving customers a layer that can switch between models depending on the task and the economics of deployment.
That matters for companies trying to expand AI use without letting costs rise unchecked. Businesses deploying assistants, search tools, and automated agents often face trade-offs between speed, quality, and spending, and some buyers see open models as one way to improve that balance.
"Enterprises are moving beyond the idea that one model should do everything," said Emrecan Dogan, Chief Product Officer at Glean.
"They want the ability to match the right model to the right task, and they need a cost-effective way to bring AI into everyday work. Our support for NVIDIA Nemotron 3 Ultra reflects that reality and gives customers a strong option as they scale AI across the enterprise," Dogan said.
NVIDIA framed the partnership similarly, focusing on the role of model selection in day-to-day business deployment. The chipmaker has been expanding the reach of its AI software and model ecosystem as it seeks to strengthen its position in enterprise generative AI beyond hardware.
"Glean is bringing NVIDIA Nemotron 3 Ultra into enterprise AI workflows where model choice, cost, and performance are critical," said Kari Briski, Vice President of Generative AI at NVIDIA.
"Together, we're helping companies deploy open models for everyday work at scale," Briski said.
Cost pressure
The addition comes as corporate buyers scrutinise the expense of generative AI more closely. Early enthusiasm for broad deployment has increasingly been matched by questions about operating costs, especially when more advanced models are used for large volumes of tasks that may not require top-tier reasoning.
Glean's approach is to allocate work across different models, using smaller or cheaper systems for some tasks while reserving more advanced models for jobs that require them. The company said Waldo handles search work that frontier models previously managed, leaving those models for tasks that need more reasoning and deeper responses.
That approach is becoming more common across the market as software companies try to show a clearer return on AI spending. Routing, orchestration, and retrieval have become as important as model access itself, particularly in enterprise settings where systems must respect internal permissions and company data boundaries.
Glean, which describes itself as a work AI platform, has focused on search, assistants, and agents linked to company data. It said the platform is designed to let organisations deploy multiple models within a secure, context-aware environment while avoiding dependence on a single provider.
Glean has been valued at USD $7.2 billion, underscoring investor interest in software groups seeking to position themselves between businesses and the growing number of AI model providers.