Datadog 推出面向 SRE 的前沿 Agent 模型 Scout 1.0

今天，我们推出 Scout 1.0——Datadog 首个专门用于调查生产环境问题的 AI 模型。Scout 为我们的 AI SRE 产品 Bits Investigation 提供支持。我们训练 Scout 的目标是降低推理成本，使调查的成本足够低，从而能够在事故生命周期更早阶段、以高出一个数量级的频率运行调查。

主要亮点包括：

• 与通用前沿模型相比，Scout 在我们的目标场景中实现了更优的性价比曲线（图 1）。Scout 1.0 medium 的表现可匹配 GPT-6.1 Sol high 和 Claude Opus 5.5 high，而预估推理成本低 3—8 倍；其性能也与 Fable medium 和 Astra medium 相当，但成本分别低 12 倍和 19 倍。通过测试时扩展（test-time scaling），Scout 1.0 xhigh 可在成本低 4 倍的情况下匹配 Fable xhigh。

• 我们利用自研的分布式强化学习基础设施构建 Scout，并广泛使用真实历史事故和遥测数据来创建高质量训练模拟环境，其中不包含任何客户数据。有关更多技术细节，请参阅我们的完整报告。

• Datadog 工程团队以及参与封闭预览的一小部分客户，已经在使用 Scout 调查生产告警。请报名加入预览，在你的环境中试用并在正式全面发布前提供反馈。

图 1：在 AlertEval——我们内部由人工专家标注的调查基准——上，调查质量与预估推理成本的对比。在中等投入模式下，Scout 的得分可与 GPT-6.1 Sol 和 Claude Opus 5.5 相当，但推理成本低 3—8 倍。最高投入模式 Scout 1.0 xhigh 实现了最先进的准确性，优于 GPT-6 Astra，并可匹配 Claude Fable 5.1 xhigh，同时将成本维持在与 Sol 和 Opus 相同的范围内。*

我们的目标是让 Bits 成为始终在线的第一响应者，在工程师介入前收集证据并识别可能原因。每次 Bits 调查都涉及提出假设、查询遥测数据，以及判断证据是否支持某个结论（图 2）。这些步骤需要重复调用模型。通过降低模型的推理成本，我们能够为团队创造更多空间，在其环境中运行调查。Scout 的效率使得将支持扩展到更多告警、并更深入地调查复杂问题成为现实，从而实现更广的覆盖范围，减轻值班工作负担，并为开发释放更多时间。

更低的推理成本还使得我们能够在更困难或影响更大的问题上投入额外计算资源。例如，Scout 1.0 xhigh 会为每起事故使用更多计算量，以提高调查质量；这种额外的调查深度可在即使最关键的事故中降低工程工作量和压力。

图 2：Bits Investigation 展示 Scout 对一起 Postgres 事故作出的诊断及其建议的恢复操作。

我们在 GLM-5.3 Flash 的基础上构建 Scout 1.0，并在调查模拟环境中通过强化学习对其进行了后训练。我们利用来自真实告警的历史事故和遥测数据，构建具有已知原因的逼真模拟环境。Scout 会练习使用基于 Bits Investigation 建模的工具来收集证据，并检验相互竞争的解释。其训练机制会奖励发现故障原因、并用证据支撑结论的能力。

我们还在开源工具基础上构建了自己的训练基础设施，具备用于大规模运行强化学习的模拟、评估和模型训练系统。

为了测试这些能力能否迁移到真实事故中，我们构建了 AlertEval。Datadog 工程师通过我们的事故响应工作流整理了数千个评估数据点。AlertEval 基于这些工作，将真实告警、归档遥测数据以及工程师编写的根本原因说明结合起来。

我们正在同步构建 Scout 和 Bits Investigation：一方面围绕 Bits 所使用的工具训练模型，另一方面根据其调查中获得的经验改进这些工具。

我们还在第三方基准 SREGym 上评估了 Scout，以测试其在 AlertEval 之外的表现。我们再次发现，Scout 1.0 medium 相对于基础模型 GLM-5.3-Flash 的性能有显著提升，并且在成本降低超过 4 倍的情况下超过 GLM-5.3。通过进一步的测试时扩展，我们发现 Scout 1.0 high 在该基准上可与最佳模型竞争，同时相较所有竞争对手仍至少保持 2 倍成本节省（相较 Astra/Fable 级别模型可节省 10 倍）。

未来，我们计划模拟更广泛的故障类型，并在持续降低推理成本的同时继续提升调查质量。我们还计划将这一方法从调查扩展到帮助团队自主修复问题，以更快恢复服务。这些进展将推动 Bits 超越第一响应者的角色，向事故期间唯一响应者发展，实现更快的问题解决并降低值班负担。

预览客户已经在其生产告警中使用由 Scout 驱动的 Bits Investigation。请报名加入预览，在你的环境中试用并在正式全面发布前提供反馈。

感谢所有帮助构建 Scout 的人员；完整作者名单和致谢请参阅我们的技术报告。

*这些分数衡量的是与专家参考解释的一致程度，采用 0—100 分量表；它们并不代表已解决事故的百分比。成本是相对于 Scout 1.0 medium 的每次调查预估推理成本。成本越低，图中位置越靠右。

Today, we’re introducing Scout 1.0, Datadog’s first AI model specialized for investigating production issues. Scout powers Bits Investigation, our AI SRE. We trained Scout to reduce inference cost so that investigations are cheap enough to run an order of magnitude more often and earlier in an incident’s lifecycle.Â

Key highlights include:

• Scout achieves a better price-performance curve on our target scenarios when compared with general-purpose frontier models (Figure 1). Scout 1.0 medium matches GPT-6.1 Sol high and Claude Opus 5.5 high at 3–8× lower estimated inference cost, and performs comparably to Fable medium and Astra medium at 12× and 19× lower cost, respectively. With test-time scaling, Scout 1.0 xhigh matches Fable xhigh at 4× lower cost. Â

• We built Scout using our in-house distributed reinforcement learning infrastructure, and relied extensively on real historical incidents and telemetry data to build high-quality training simulations that don’t contain any customer data. For more technical details, see our full report.

• Datadog engineering teams and a small group of customers in a closed preview are already using Scout to investigate production alerts. Sign up for the preview to try it in your environment and provide feedback before general availability.

Figure 1: Investigation quality versus estimated inference cost on AlertEval, our internal human-expert-annotated investigation benchmark. In its medium-effort mode, Scout achieves a score comparable to GPT-6.1 Sol and Claude Opus 5.5, at 3×–8× lower inference cost. Our highest-effort mode, Scout 1.0 xhigh, achieves state-of-the-art accuracy that outperforms GPT-6 Astra and matches Claude Fable 5.1 xhigh while keeping costs in the same range as Sol and Opus.*

Our goal is for Bits to serve as an always-on first responder, gathering evidence and identifying likely causes before an engineer steps in. Each Bits investigation involves forming hypotheses, querying telemetry, and deciding whether the evidence supports a conclusion (Figure 2). Those steps require repeated model calls. By reducing a model’s inference cost, we can give teams more room to run investigations across their environments. Scout’s efficiency makes it practical to extend support to more alerts and investigate complex issues in greater depth, enabling greater breadth of coverage, reducing the burden of on-call work, and freeing more time for development.

Lower inference cost also makes it practical to spend additional computation on harder or higher-impact problems. For instance, Scout 1.0 xhigh uses more computation per incident to improve investigation quality, and this added investigative depth reduces engineering workload and pressure during even the highest-stakes incidents.

Figure 2: Bits Investigation presents Scout’s diagnosis and proposed recovery actions for a Postgres incident.

We built Scout 1.0 on GLM-5.3 Flash and post-trained it using reinforcement learning in investigation simulations. We use historical incidents and telemetry from real alerts to build realistic simulations with known causes. Scout practices using tools modeled on Bits Investigation to gather evidence and test competing explanations. Its training rewards finding the cause and supporting its conclusions with evidence.

We also built our own training infrastructure on top of open source tools, with the simulation, evaluation, and model-training systems needed to run reinforcement learning at scale.

To test whether those skills transfer to real incidents, we built AlertEval. Datadog engineers have curated thousands of evaluation points through our incident-response workflow. AlertEval draws on that work, combining real alerts, archived telemetry, and root cause explanations written by engineers.

We are building Scout and Bits Investigation together, training the model for the tools Bits uses and improving those tools based on what we learn from its investigations.

We have also evaluated Scout on SREGym, a third-party benchmark, to test its performance outside AlertEval. We again have found that Scout 1.0 medium significantly improves upon the base model, GLM-5.3-Flash, performance, surpassing GLM-5.3 at a more than 4× reduction in cost. Through additional test time scaling, we found that Scout 1.0 high is competitive with the best models on this benchmark, while still maintaining at least 2× cost savings compared to all competitors (10× savings compared to Astra/Fable-class models).

Moving forward, we plan to simulate a wider range of failures and keep improving investigation quality while lowering inference cost. We also plan to extend this approach beyond investigation to help teams autonomously remediate issues to restore service faster. Together, these advances move Bits beyond its role as a first responder and toward becoming the only responder during an incident, resolving issues faster and reducing on-call burden.

Preview customers are already using Bits Investigation powered by Scout on their production alerts. Sign up for the preview to try it in your environment and provide feedback before general availability.

Thanks to everyone who helped build Scout; see our technical report for the full author list and acknowledgments.

*The scores measure agreement with expert reference explanations on a 0–100 scale; they are not percentages of incidents solved. Costs are estimated inference costs per investigation, relative to Scout 1.0 medium. Lower cost appears to the right.