Carnegie Mellon University will host the inaugural ProtoStar Challenge from Nov. 2–13, 2026. Its first event, Climate Risk Intelligence, is designed to test how AI can use historical weather records from 210 U.S. cities for climate-risk analysis and adaptation planning. CMU announced the challenge on Oct. 29.
What the challenge is designed to examine
CMU says the challenge will draw on historical temperature and precipitation records for 210 U.S. cities. Hugging Face’s launch description names four planned measures: annual average temperature, annual precipitation, the hottest day of the year, and cooling degree days, a proxy for air-conditioning demand. Hugging Face outlined the proposed task.
The launch description proposes comparing AI forecasts with two straightforward baselines: historical averages and linear trends. Those comparisons are intended to assess how well forecasts match observed weather. The challenge is a planned evaluation, rather than a reported demonstration of improved climate forecasts.
Who is involved and when results are planned
CMU is hosting the challenge with Hugging Face and university partners through the Open Forum for AI (OFAI). The collaboration also includes Purdue University, Georgia Institute of Technology, Arizona State University, George Washington University, the Open Research Community Accelerator (ORCA), and Astera Institute.
Hugging Face said final forecast results were planned for early 2027, after complete 2026 weather records became available.
