VMware Adds Google, Nvidia AI Models To VCF; Boosts Tanzu Security To Drive AI Adoption
‘The more that VMware puts together easier bundling and makes it easier for our customers to operationalize AI—the more we see less barriers for adoption,’ says Bob Keblusek, CTO at VMware partner Sentinel Technologies.
VMware has added many of the world’s most popular AI models to its VMware Cloud Foundation (VCF) private cloud platform to enable partners to bring more models on-premises and provide model-as-a-service for customers.
Some of the newest AI models from Google, Nvidia, NEC, Alibaba Cloud and Z.ai have now been tested and validated to run on VCF.
“The more that VMware puts together easier bundling and makes it easier for our customers to operationalize AI—the more we see less barriers for adoption,” said Bob Keblusek, CTO at VMware partner Sentinel Technologies.
Keblusek said VMware is giving clients a clear path to data sovereignty and cost-effective AI at scale, with leading models securely available and delivered as a service via VCF.
“For Sentinel’s part, we have cost optimization dashboards and some FinOps basically around tokenization because it starts to really accelerate as you're adopting AI, especially if you’re using frontier models,” said Keblusek. “So a lot of the use cases can be localized.”
Google Gemma 4, Nvidia Nemotron 3 And Other AI Models Now On VCF
Two of the most popular AI models now validated on VCF include Google’s Gemma 4 and Nvidia’s Nemotron 3.
Gemma 4 is Google’s open-source, open-weight multimodal model family, purpose-built for developers and researchers—enabling enterprises to build and deploy autonomous AI agents.
Nvidia’s Nemotron 3 family of open, multimodal models aim to deliver accuracy and efficiency to help agents complete tasks faster. VMware said by combining hybrid Mamba-Transformer MoE architecture, 1 million context and multi-environment reinforcement learning, Nemotron 3 enables long-running agentic workflows across enterprise applications.
Another new AI model for VCF includes Z.ai’s GLM 5.2 open-source General Language Model for deploying coding and reasoning agents locally for multistep autonomous workflows with data sovereignty.
The final two new AI models for VCF include NEC’s cotomi model and Alibaba’s Qwen 3.8-27B open-weight model.
Over 150 Open-Source AI Models Now On VCF
VMware said customers now have the ability to run more than 150 open-source models on VCF.
Keblusek said VMware is committed to giving customers a broad set of AI models for their on-premises infrastructure, all validated on VCF.
“It’s good to be able to localize and have control over that, which you don’t have in some cases, depending on the product,” he said, adding that VCF can run inference workloads, agentic applications, containerized services and traditional VMs together, eliminating the need to manage separate stacks.
VMware Boosts Tanzu With New Security Innovation
The new AI models on VCF were unveiled this week at VMware Explore 2026 in Las Vegas.
Another huge announcement was that VMware launched new AI-ready data foundations for the VMware Tanzu Platform. The update provides an end-to-end framework enabling enterprises to transition safely from initial AI pilots to fully production-ready AI agents inside their own secure private clouds.
New capabilities included hardened agent sandboxes that enforced a “deny-by-default" security containment model that isolates credentials, helping to prevent prompt injection attacks and unauthorized network access.
“With agentic AI, it’s about how do you govern that, control that, orchestrate that, which are very large topics for us with our customers who are adopting agentic—it’s a large conversation that many vendors are taking a swing at with different approaches to handling it,” said Keblusek.
“Seeing VMware go after this with the Tanzu platform makes a ton of sense. If we’re able to add agentic security so that our customers can adopt agentic and feel comfortable about it, run it in their own data centers or in a VMware secure enclave or cloud—that is all very welcomed as security is the top concern when talking to customers about AI adoption,” he said.
Another new feature is the AI-ready data foundations that process structured and unstructured enterprise data on-site, delivering high-precision context to AI agents to improve accuracy, reduce hallucinations and lower token costs.
Other new innovations include an out-of-the-box developer harness that accelerates build times with preapproved skills, step-by-step workflows, human-in-the-loop controls and integrated memory services. Lastly, VMware launched a curated marketplace that operates a centralized catalog where developers and agents can safely discover and connect to vetted AI models, tools and data products.
“We need to really put the controls in place to enforce policies, and it starts with monitoring, which is what VMware is doing,” Keblusek said. “It starts with being able to build that secure, trustworthy infrastructure, and then we’ll see AI adopted at even higher rates than we see today thanks to VMware innovation.”