Open & Local AI Language Models: Part of a More Sustainable Future for Scientific Computing

By DSE September 8, 2026

Researchers and local communities alike are growing concerned about the impacts of using Large Language Models (LLMs) due to the extensive impacts these AI models and the data centers they rely on have on the environment – particularly on air and water quality. At the same time, alternative approaches to AI energy demands are already unfolding much closer to home (on our own laptops, in fact!). 

 

Here at the Schmidt Center for Data Science & Environment (DSE), we share concerns about the impacts of LLM use on the environment. That’s why we and many researchers are increasingly pivoting to using and developing open source and local AI models. These models tend to be far more energy efficient than their commercial counterparts and run for free. Let’s explore the benefits of open source and local AI models, and the ways in which these tools are already advancing scientific computing on a more sustainable path.

 

What are Open & Local AI Models?

“AI” is often used as a general term for all kinds of language models, and the reality is there is a large spectrum of tools that fall under this umbrella. At the same time, many of us consider all types of AI to be entirely reliant on massive data centers to function. But that’s not true! Not all AI requires the use of data centers.

 

For example, open and local AI models are open source, which means that they are owned by the user (rather than a commercial entity). Anyone can view, use, modify, and distribute these tools. They can also download and run locally on a user’s machine using the power from their laptop or desktop – no data center required.

 

This graphic below depicts the key benefits of using open and local AI models in science: robust research, energy efficiency, increased privacy and reliability, freedom of use, and low cost. 

 

infographic depicting benefits of local language models
Using local language models brings many benefits for scientists, programmers, and our respective fields more broadly. Read a detailed image description.

 

How are Open & Local Models different from tools ChatGPT and Claude?

Because open models and local models run on a user’s computer machine, these models may use 100 times less energy than commercial LLMs. At the same time, open and local AI models are more often performing at pace with leading commercial models such as ChatGPT and Claude. Read an interview with DSE faculty advisors in Berkeley News about how we can reduce reliance on data centers using open and local models. 

 

DSE does use, and will continue to use, commercial LLMs to meet the needs of our partners and collaborators. From our perspective it’s important to pick the right tool for the right use case. In this way, both commercial and open/local models will continue to be relevant in our work long term.

 

How is DSE using Open & Local Models?

While using open and local models does require some technical expertise, they offer immense opportunities for scale and impact in science and beyond. Since DSE was founded in 2022, we have co-developed nine open source, decision-support tools for policymakers and scientists. We are increasingly pivoting to developing open and local models for stakeholders, including Kigali Sim for the United Nations (UN) to help reduce global greenhouse gas pollution and the Geospatial LLM-Enabled Navigator (GLEN) High Seas Explorer to establish global new marine protected areas. 

 

GLEN compiles and visualizes data on existing protected areas, ocean floor geology, biodiversity, and global fisheries activity and links this data to UN policy text. With the chatbot, users can ask questions like, “if I created a new marine protected area that was 100 kilometers wide in the high seas right next to my country’s border, how many undersea mountains and how many whale migration pathways would be included?” and will receive scientifically-rigorous answers in minutes. Read a commentary in Nature from DSE on how GLEN exemplifies the many benefits of open and local AI models. And visit our GLEN project page to see many more case studies for this tool.

 

screenshot from GLEN tool
A screenshot from the GLEN High Seas Explorer. Ecologically or Biologically Significant Areas (EBSAs) can inform new, large-scale Marine Protected Areas via the new United Nations High Seas Treaty. When the following question was posed to the High Seas Explorer: “Do High Seas EBSAs contain, on average, a greater number of threatened species compared to other High Seas regions?” The resulting spatial analysis revealed that High Seas EBSAs are home to a greater number of threatened species compared to other High Seas areas. The query’s corresponding spatial layer uses a color scale ranging from yellow (fewer threatened species) to red (greater number of threatened species).

 

What’s Next?

Scientists have led the adoption of open source tools from the internet to programming languages like Python. Widespread use of open source and local AI models will require more investment from academic institutions, but it’s critical that our field as a whole pursue this more sustainable and ethical path in computing. As we increase our investments in developing more open and local models at DSE, we hope to see more researchers and institutions do the same. 

Program