Analysis: recent open weight models lag frontier closed models' cyber capabilities by 4 to 7 months, a narrower gap than the 6 to 10 months through most of 2025 (AI Security Institute)
.png)
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- AISI has tracked the cyber capabilities of frontier AI models since 2023 as part of its ongoing evaluation program, according to the institute's analysis.
- The lag between open weight models and frontier closed models on cyber tasks narrowed from 6 to 10 months through most of 2025 to 4 to 7 months in the most recent assessment.
Why it matters: A 4-to-7-month lag means open-weight models are reaching frontier-level cyber performance on a faster timeline than the 6-to-10-month pace seen through most of 2025, narrowing the window between when capability emerges in closed frontier systems and when it becomes available in openly distributed weights.
