Snorkel AI Names First Open Benchmarks Grant Projects

Snorkel AI announced in a press release the first group of projects supported by Open Benchmarks Grants, a $3 million program for open source datasets, benchmarks, and AI evaluation research. The program launched in February 2026 and provides selected teams with funding, data support, research and engineering collaboration, and platform resources.

The first supported projects include Frontier-Bench, formerly Terminal-Bench 3.0, with Laude Institute and the Harbor community, and Agents' Last Exam, with UC Berkeley RDI and the RDI Foundation. Other projects include OSWorld 2.0 with XLANG Lab, Continual Learning Bench with UC Berkeley SkyLab and the University of Wisconsin Madison, SlopCode Bench with the University of Wisconsin Madison, and Terminal-Bench 2.1 with Stanford University, Laude Institute, and the Harbor community.

Snorkel AI also said Terminal-Bench Science is in development to extend the Terminal-Bench framework to computational research workflows in life, physical, earth, and mathematical sciences. Separately, the company led Senior SWE-Bench with research teams at Princeton University and the University of Wisconsin Madison to evaluate coding agents on senior level engineering tasks.