Updates from Snorkel AI
Expert data development and environments for frontier AI models and agents.
- Terminal-Bench 4.1 hardens the verifiers behind the scores (Project update · November 9, 2026)
Snorkel AI released Terminal-Bench 4.1, an updated evaluation framework for terminal agents. The new version features hardened verifiers, reproducibility improvements, and a new dataset designed for… - Agents' Last Exam (ALE) Leaderboard: v1.1 Results (Project update · October 26, 2026)
Snorkel AI introduced the Agents' Last Exam v1.1 leaderboard, evaluating AI agents on 122 long-horizon professional tasks. The benchmark measures performance across 55 sub-industries using… - Frontier AI is Accelerating. Open Benchmarks Need to Keep Up. (Company update · October 7, 2026)
The initiative supports researchers and domain experts in developing robust evaluation methods for frontier AI. - Snorkel AI, Princeton, and UW-Madison researchers release Senior SWE-bench (Project update · July 1, 2026)
Snorkel AI and researchers from Princeton and UW-Madison released Senior SWE-bench, a new evaluation framework for coding agents. The benchmark focuses on senior-level software engineering tasks to…