The investment targets the risk of "benchmaxxing," where frontier models are optimized for specific tests rather than actual intelligence, rendering metrics static or too simple. Snorkel AI has already collaborated with the developers of Terminal-Bench, ARC-AGI-3, and OSWorld 2.0, and the company noted that its funded benchmarks are currently featured on model cards from every major frontier lab to index progress and alignment.
Tuesday, October 20, 2026
Story
Snorkel AI Increases Open AI Benchmark Funding 10x to $30M
AI13 days ago
A series of new initiatives from Snorkel AI aims to prevent AI benchmarks from becoming obsolete as large-scale models evolve. The company has increased its financial commitment to the Open Benchmarks Grants tenfold to $30 million to support a more diverse and continuously updated ecosystem of open measurement tools. This expansion includes the launch of an Open Benchmarks Red Team to test and strengthen existing benchmarks and the introduction of the Snorkel Research Fellowship for contributors developing new evaluation methods.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.