Dive Brief:
- Google released the open source AI model family Gemma 4 on Thursday, which was designed to help with reasoning, agentic work and code generation. The company joins other large providers who are aiming to serve open source models to enterprise clients.
- The Gemma 4 models are being released under the Apache 2.0 license, and come in several model weights, designed for different hardware use cases including Android devices, laptop GPUs, workstations and accelerators.
- Open source models are gaining enterprise traction due to being cost-efficient, customizable and adaptable to a multi-modal approach. More than three-quarters of companies reported using two or more LLM families, including a mix of closed and open-source models, a 2026 Databricks report found.
Dive Insight:
Enterprises are leaning toward a mix of proprietary AI technology and open source models, keeping cost and latency in mind.
Open source models can be more easily tailored to specific business use cases and allow for more control over data and infrastructure, which make them attractive to industries with strict data privacy requirements. Gemma 4 lets developers run smaller AI tasks locally and offline, another added benefit for enterprises concerned with security.
Still, open source models bring their own challenges, including a lack of built-in safety guardrails often found in closed systems. It can be harder to ensure the data a model was trained on meets an organization’s standards, said Chirag Dekate, VP analyst at Gartner.
While open source models like Gemma 4 allow flexibility with tasks, it’s not a one-size-fits-all model for a company’s every need, Dekate said.
Enterprises shouldn’t rush to apply open source models to all of their AI use cases. Creating a complex financial model could likely still require a proprietary model with guardrails and advanced capabilities, Dekate said.
“CIOs should look at this as a portfolio where they create a mix of open models as well as a handful of proprietary models, and create the right balance for their evolving use case,” Dekate said. “Here, agility matters more than dogmas.”
Supply chain durability is one of the biggest considerations for enterprises when mixing open-source and proprietary models, Dekate said. He cautioned tech leaders to consider if the open source model will stay open forever, as China’s Alibaba shifted its popular Qwen model family to proprietary earlier this year.
The durability of open source models is key, given AI’s role as a core intelligence layer that powers the enterprise and the products it creates, Dekate said.
