Gemini 3 beaks OpenAI's long-standing lead in SRE tasks
A major shift just hit SRE-focused AI. Gemini 3 Pro edged out OpenAI’s models and outperformed them across every single SRE task we tested.
In this Rootly AI Labs episode, Sylvain Kalache and Laurence Liang break down:
- Why Gemini 3 Pro came out on top for compute, storage, and networking actions
- How the SRE-skills-bench benchmark works and what it actually measures
- Real examples of SRE-type tasks
- What these results mean for SREs, platform engineers, and on-call teams in the near future