Study: Frontier AI Time Horizon Reaches 14 Hours at 50M Token Window
tomekkorbak · x · 2026-07-03
The AI Safety Institute found that the time horizon of frontier AI evaluated under a 50M token context window is about 60% shorter than under a 2.5M token window. Specific data: The frontier time horizon is 2 hours under the 2.5M token window, and 14 hours under the 50M token window.
This finding indicates that the true trends of AI capabilities can only be observed when given a sufficiently large context budget—the choice of evaluation window fundamentally affects the conclusions drawn about the frontier level of AI.
Related event: Frontier AI Task Duration Doubles, Slowed by Budget Constraints(3 posts)→
More from Safety
- Meta accused of letting AI-generated fake doctors spread health advice for traffic — GaryMarcus · 2026-07-27
- ExploitGym may have only 60–70% solvable tasks, fueling the OpenAI cheating debate — max_paperclips · 2026-07-27
- Shared AI artifacts are being indexed and exposing sensitive company data — niloofar_mire · 2026-07-27
- Post-Hugging Face, labs may stop running rigorous dangerous-capability evals — Miles_Brundage · 2026-07-27
- Open models may beat closed ones for cyber defense, researchers argue as Kimi K3 impresses — eliebakouch · 2026-07-27
- India’s AI policy is favoring compute and foundation models over frontline health workers — Paimaamu · 2026-07-27