AI-300 exam format: questions, score and cost
AI-300 gives you 120 minutes, requires a scaled score of 700 out of 1000 to pass, is offered in English only, and is priced per country at the booking step. Microsoft does not publish the number of questions. Here is everything else it publishes, and what it means for your preparation.
| Exam title | Operationalizing Machine Learning and Generative AI Solutions |
| Certification | Microsoft Certified: Machine Learning Operations Engineer Associate |
| Level | Intermediate (associate) |
| Duration | 120 minutes |
| Passing score | 700 on a scale of 1–1000 |
| Number of questions | Not published by Microsoft |
| Languages | English |
| Delivery | Pearson VUE, proctored |
| Price | Set per country or region; shown at booking |
| Renewal | Annually, free, on Microsoft Learn |
| Replaces | DP-100 (Azure Data Scientist Associate), retired June 1, 2026 |
The five domains
| Domain | Weight |
|---|---|
| Design and implement an MLOps infrastructure | 15–20% |
| Implement machine learning model lifecycle and operations | 25–30% |
| Design and implement a GenAIOps infrastructure | 20–25% |
| Implement generative AI quality assurance and observability | 10–15% |
| Optimize generative AI systems and model performance | 10–15% |
Read the table as two halves. The first two domains are about Azure Machine Learning and are worth 40–50% together. The last three are about generative AI in Microsoft Foundry and are worth 40–55%. You cannot pass on one half alone.
The model lifecycle domain is the single largest, and it has the most sub-objectives: training orchestration, registration, deployment and monitoring each get their own section. The two 10–15% domains are small, but they are also the least familiar to most candidates. A few questions on evaluation metrics or hybrid search can decide a borderline result.
From beta to live
Microsoft announced AI-300 in March 2026 and opened it as a beta exam, with general availability planned for May 2026. Beta candidates receive their result only after the exam goes live and rescoring is complete. The certification page no longer labels the exam as beta.
The study guide is dated February 2026. Check it again before you book: Microsoft updates outlines with little notice, and the English exam changes first.
Practice assessment and sandbox
Microsoft’s free practice assessment for AI-300 is available, but not on the usual Microsoft Learn page. It lives on AI Skills Navigator, and you must be signed in there to launch it. It is the best free indication of question style and difficulty.
The exam sandbox shows the interface and question types without testing any content. It is worth ten minutes before your first Microsoft exam.
Question types
Expect the usual Microsoft mix:
- Single-answer and multiple-answer multiple choice
- Drag-and-drop ordering, for example the steps to register an MLflow model and deploy it to an online endpoint
- Dropdown completion of YAML, CLI commands, Bicep or Python snippets
- Yes/no statement series about one scenario, where you usually cannot go back once answered
- Case studies with several questions sharing one long scenario
Microsoft notes the exam may include interactive components. The outline names Bicep, the Azure CLI, GitHub Actions and Python explicitly, so expect to read snippets and spot the missing or wrong line rather than just pick a service.
Who Microsoft writes it for
The audience profile asks for a data science background with Python, plus an entry-level understanding of DevOps: GitHub Actions and working on the command line. It names four toolsets: Azure Machine Learning, Microsoft Foundry, GitHub Actions and infrastructure as code with Bicep and the Azure CLI.
“Entry-level DevOps” matters. You will not be asked to design a complex release process from scratch, but you do need to know how a workflow authenticates to Azure, how a Bicep file deploys a workspace, and why a model is promoted rather than retrained per environment.
The passing score is scaled
700 of 1000 is not 70% correct. Microsoft scales scores and weights questions differently, and does not publish the conversion. Aim for a consistent 80% or more on realistic practice material rather than working out a margin.
Preview features
Most questions cover generally available features, but Microsoft states that preview features may appear if they are commonly used. That matters for AI-300, because Foundry, its evaluation tooling and its monitoring features move fast. Study from the current documentation, and treat anything written before the Azure AI Foundry to Microsoft Foundry rename with suspicion.
Language and extra time
AI-300 is currently offered in English only. If English is not your first language, you can request an additional 30 minutes. Arrange that before you book, not on the day.
Retakes and renewal
If you fail, you can retake after 24 hours. Later retakes carry longer waiting periods under Microsoft’s retake policy. Once you pass, the certification is valid for one year and renews through a free online assessment on Microsoft Learn.
For what to do on the day itself, see the exam day guide.