CursorBench 4.0 Rolls Out Harder Long-Horizon Tasks, Model Scores Drop Across the Board

CursorBench has been upgraded to version 4.0 with new long-horizon tasks testing instruction following and sustained work in complex projects, making it significantly harder and causing scores to drop across all models.

2026-09-11 ~ 2026-09-11 · 2 related posts

• CursorBench 4.0 rolls out with harder, longer-horizon coding tasks, scores drop — StringChaos · 2026-09-11
• CursorBench 4.0 launches; Muse Spark 1.3 matches Sol at under 40% the cost — jyangballin · 2026-09-11