Post

Introducing Scout 1.0: A new SRE model powering Bits AI with frontier performance at a fraction of the cost

Datadog introduced Scout 1.0, a specialized AI model for automated production incident investigation, achieving 3–8× lower inference costs than frontier models (GPT-6.1 Sol, Claude Opus 5.5) while matching investigation accuracy. Scout was trained via reinforcement learning on simulations built from real historical incidents and telemetry without customer data, evaluated on AlertEval (Datadog’s curated incident benchmark) and third-party SREGym. The model powers Bits Investigation, which autonomously gathers evidence, forms hypotheses about root causes, and tests competing explanations; reduced inference cost enables broader alert coverage and deeper investigation of high-impact incidents, reducing on-call burden.

Read original article

This post is licensed under CC BY 4.0 by the author.