Capital · United States

UChicago Receives $10M Grant to Build Open-Source AI Weather Forecasting Platform

Credited to HPCwire · hpcwire.com · 1h ago

The grant will allow the UChicago team to build new AI weather forecasting frameworks.

CHICAGO, Nov. 6, 2026 — If you had the resources, how would you use AI to solve humanity’s hardest problems? That question is at the heart of a $10 million grant the Laude Institute awarded to a University of Chicago team building AI weather forecasting models and supporting a software platform for people who depend on them for their health and livelihoods.

The PI Team from left to right: Prof. Ian Foster, Prof. Rebecca Willett, Prof. Michael Kremer, and Prof. Pedram Hassanzadeh. Photo Credit: Victoria Smith.

In just two seasons, governments in South Asia and Africa have used AI forecasts developed with the UChicago team to give more than 50 million smallholder farmers life-changing information: when the rainy season will arrive. The Laude funding will make that process faster, simpler, and more accurate, and extend it to new regions and new weather phenomena.

For example, a farmer whose livelihood depends on knowing when the monsoon season will begin must decide whether to plant now or wait. However, farmers in tropical regions have had less accurate and less actionable forecasts. Wealthy nations developed modern, physics-based weather forecasting for mid-latitude climates, using dense observation networks and supercomputers that most tropical agencies lack. And forecasts are typically built to predict general conditions like temperature and precipitation, not the specific events, such as when the rains will arrive, that drive farmers’ decisions.

The UChicago team is led by scientific AI expert Prof. Rebecca Willett, Worah Family Professor of Statistics and Computer Science in the Wallman Society of Fellows; AI for atmospheric physics expert Prof. Pedram Hassanzadeh, Associate Professor in Geophysical Sciences; platform scaling expert Prof. Ian Foster, Arthur Holly Compton Distinguished Service Professor of Computer Science; and development economics expert Prof. Michael Kremer, Nobel laureate and University Professor in Economics and the College and the Harris School of Public Policy.

This team is working to close the gap between mid-latitude and tropical forecasting abilities through AI-powered systems that work with sparse observational data and modest computing resources. Working at the frontier of AI weather modeling, the group has developed methods that take the best existing AI and physics-based forecasts, blend them, and tune them to the specific human-centered decisions that farmers, health officials, and emergency managers actually face. Over the last six months, the UChicago team and collaborators have cut the time needed to benchmark and adapt forecasting models from four to five months of expert labor to a single day of expert review.

Willett said, “The leading AI weather models are evaluated on whether they accurately predict global temperature and precipitation. That’s useful if you’re a meteorologist. But it tells you almost nothing about whether a model will correctly predict monsoon onset for a farmer in West Java or a dangerous heat wave for a public health official in Addis Ababa. We built our platform around a different question: not ‘how accurate is the model’ but ‘does this forecast help a specific person make a better decision in a specific place.’”

With Laude Institute support, the team will build new AI weather forecasting frameworks embedded in the UChicago–Laude AI Almanac, a free, open-source platform that will let any meteorological agency, government, or NGO generate, evaluate, and deploy AI weather forecasts tuned to their own geography, observations, decision needs, and knowledge of local weather dynamics without requiring specialized technical expertise, supercomputing infrastructure, or dense observational networks.

“This platform will significantly accelerate the translation of the AI-driven weather forecasting revolution into enhanced capabilities of the meteorological agencies that need it most, so they can give high-quality, actionable information to the people they serve,” said Hassanzadeh.

Early deployments have already shown what’s possible using AI weather forecasts. From 2024 to 2026, the UChicago Human-Centered Weather Forecasts initiative supported government agricultural ministries in producing AI weather forecasts that were sent via text message to tens of millions of farmers in Africa and South Asia. Those AI forecasts correctly predicted an unusual stall in the seasonal rains, providing an early warning to farmers. Farmers who received the AI-powered forecast on their phones could make informed planting decisions, rather than planting too early and losing their crop. The Laude grant will help turn this work into a global platform covering many kinds of weather and climate hazards.

Development banks and philanthropies such as the OpenAI Foundation and the Gates Foundation are investing heavily in delivering accurate weather forecasts and decision-support tools to farmers, public health workers, and government officials in low- and middle-income countries, with a goal of reaching hundreds of millions of people within the next several years.

The UChicago team works closely with these organizations, as well as AIM for Scale, Precision Development, and UC Berkeley, on forecast delivery. The work supported by the Laude Institute will strengthen the AI and computational backbone behind those efforts with new AI methods, trusted evaluation focused on local community needs, and a scalable, open platform that expands access and increases forecast reliability, without major investments in computing infrastructure or software development. The platform will support not only agriculture but also public health and warnings of compound climate extremes.

“Reaching hundreds of millions of people requires infrastructure that can scale across countries, datasets, and weather hazards,” said Foster. “With Globus, a non-profit research data management platform, we have learned how to make large-scale research computing reliable and accessible. We will draw on that experience to help meteorological agencies adapt and deliver forecasts for their communities without having to build and maintain their own computing infrastructure.”

A companion AIM for Scale training program, co-led by UChicago and the United Arab Emirates’ Mohamed bin Zayed University of Artificial Intelligence, is already preparing government meteorologists from 14 countries to use these tools, with the goal of reaching 30 countries by 2027

“If this works for weather, it is a model for AI for public health, for education, and for many other areas where the people who need these tools most cannot pay for them,” said Kremer. “Philanthropy can step up where the market cannot. That is what makes this moment extraordinary.”

Built by and for impact researchers, the Laude Institute operates at the intersection of academia and industry to catalyze research that reaches people’s hands. It convenes, resources, and amplifies computer scientists who ship their research and impact the lives of billions. With carefully chosen targets, AI research has an opportunity to move all of humanity forward, and the UChicago-Laude AI Almanac aims to contribute significantly to that momentum.

The project team includes Adam Marchakitus, Dr. Alexander Wikner, Andrew Dennehy, Raphael Rossellini, Dr. Rajat Masiwal, Rathan Sudheer, Dr. Bo Dong, Dr. Casey Keel, Dr. Colin Aitken, Hayden Holbrook, Dr. Katie Kowal, Terrina Govender, Tiana Pyer-Pereira, and Dr. Victoria Flores.

Models & LabsUnited States

Ripple-65B Preview delivers strong agentic coding results as a fully Apache 2.0 open US model

Credited to The Decoder · the-decoder.com3h ago

Useful

AI chip design firm Ricursive Intelligence has trained an open language model. Ripple-65B Preview was pretrained on 12 trillion tokens and, according to the company, achieved a score of 80.52 on the TerminalBench 2.1 coding benchmark, the best result among open US models. Ripple partly uses Terminus 2 and partly OpenCode as its harness. Other benchmark results show that Ripple does not match the also newly released Inkling 2 Small, but Ripple is likely easier to deploy on users' own hardware. According to Ricursive, total development and training costs amounted to just $32 million, covering all research and development expenses.

The model uses a dense Transformer architecture with 65 billion parameters instead of the more common mixture-of-experts approach and is optimized for Nvidia hardware. Ripple-65B Preview is available on Hugging Face under the Apache 2.0 license.

AI News Without the Hype – Curated by Humans

Original · The Decoder

Front