AI-300

AI-300: Machine Learning Operations Engineer

Everything for Microsoft exam AI-300: format, passing score, the five MLOps and GenAIOps skill areas, a 4-week study plan and free sample questions.

Duration
120 min
Passing score
700 / 1000
Replaces
DP-100
Languages
English

AI-300, Operationalizing Machine Learning and Generative AI Solutions, is Microsoft’s exam for the engineers who get models into production and keep them there. Passing it earns the Microsoft Certified: Machine Learning Operations Engineer Associate credential.

It replaces DP-100. The Azure Data Scientist Associate certification was retired on June 1, 2026, and Microsoft positions AI-300 as its successor. The emphasis moved from training a good model to running one: infrastructure as code, CI/CD with GitHub Actions, deployment and rollback, drift detection, and the same discipline applied to generative AI in Microsoft Foundry.

Microsoft calls the combination AI operations: MLOps for traditional models in Azure Machine Learning, and GenAIOps for foundation models, prompts, evaluations and agents in Foundry. The exam tests both halves.

What the exam covers

Skill areaWeight
Design and implement an MLOps infrastructure15–20%
Implement machine learning model lifecycle and operations25–30%
Design and implement a GenAIOps infrastructure20–25%
Implement generative AI quality assurance and observability10–15%
Optimize generative AI systems and model performance10–15%

The model lifecycle domain is the largest, and it is the one closest to DP-100. The three generative AI domains together are worth 40–55%, which is where most former DP-100 candidates have the most to learn.

Where to start

Read the exam format for the mechanics, then what changed from DP-100 if you studied for the old exam. The study plan fits in four weeks if you already train models in Azure Machine Learning.

AI-300 guides

  1. AI-300 exam format: questions, score and costThe AI-300 format: 120 minutes, passing score 700, English only, question types, pricing, and what the five MLOps and GenAIOps weights mean.
  2. AI-300 vs DP-100: what changed?DP-100 retired on 1 June 2026 and AI-300 replaced it. What carries over from the data scientist exam, what was dropped, and the new GenAIOps content.
  3. AI-300 study plan: pass in 4 weeksA four-week AI-300 study plan: IaC and workspaces, the model lifecycle, Foundry GenAIOps, then evaluation and RAG tuning, with a hands-on lab each week.
  4. Free AI-300 sample questions with answersFive free AI-300 practice questions on GitHub Actions, endpoint rollout, drift, Foundry deployments and groundedness, each with an explained answer.
  5. AI-300 MLOps infrastructure explainedThe AI-300 MLOps infrastructure domain: Azure Machine Learning workspaces, compute, environments, registries, Bicep, GitHub Actions and network isolation.
  6. AI-300 ML model lifecycle explainedThe largest AI-300 domain: MLflow tracking, sweeps, pipelines, model registration, online and batch endpoints, safe rollout and drift monitoring.
  7. AI-300 GenAIOps infrastructure explainedThe AI-300 GenAIOps domain: Foundry resources and projects, identity and private networking, model deployment types, PTUs and prompts versioned in Git.
  8. AI-300 GenAI quality and observabilityThe AI-300 evaluation and observability domain: test datasets, groundedness and relevance, safety evaluators, tracing and token cost in Foundry.
  9. AI-300 GenAI optimization: RAG and tuningThe AI-300 optimization domain: chunk size, similarity thresholds, hybrid search, embedding models, fine-tuning methods and synthetic training data.
  10. Is AI-300 hard? Exam difficulty explainedHow hard AI-300 really is: why its two halves, the DevOps tooling, the generative AI evaluation content and the English-only exam shape the difficulty.
  11. Is AI-300 worth it? Who should take itIs AI-300 worth taking? Who benefits, what it costs, and why the retirement of DP-100 makes it Microsoft's associate exam for MLOps and GenAIOps work.
  12. Free AI-300 practice test: 20 questionsA free 20-question AI-300 practice test on Azure Machine Learning, GitHub Actions, endpoints, Foundry, evaluation and RAG, with explained answers.
  13. AI-300 study resourcesWhat to study for AI-300, which DP-100 material still helps, the official practice assessment, and six labs that cover most of the outline.
  14. AI-300 cheat sheetA one-page AI-300 reference: the five skill areas, which Azure Machine Learning or Foundry feature answers each requirement, evaluators and common traps.