The short version
Take AI-200 if you build application back ends on Azure and are adding AI features: containers, vector databases, messaging, Functions, secrets, and monitoring. Take AI-300 if you run machine learning and generative AI workloads: pipelines, deployment, evaluation, and tuning. They are separate credentials with no listed prerequisite exam. Take both only if your job spans both sides. Facts below come from Microsoft Learn pages checked on October 22, 2026.
Side by side
| Item | AI-200 | AI-300 |
|---|---|---|
| Full title | Developing AI Cloud Solutions on Azure | Operationalizing Machine Learning and Generative AI Solutions |
| Credential | Azure AI Cloud Developer Associate | Machine Learning Operations Engineer Associate |
| Level | Intermediate | Intermediate |
| Role listed | Developer | AI Engineer |
| Audience in one line | Back-end and cloud developers building AI solutions | People setting up and running MLOps and GenAIOps |
| Exam time | 120 minutes | 120 minutes |
| Passing score | 700 of 1,000 | 700 of 1,000 |
| Question count | Not published | Not published |
| Delivery | Proctored, Pearson VUE, may include interactive components | Same wording |
| Languages | 13 | English |
| Domains | 4 | 5 |
| Prerequisite exam | None listed | None listed |
| Renewal | Annual, free assessment | Annual, free assessment |
| Retakes | 24 hours, then 14 days, five tries in 12 months | Same policy |
Because the format and policies match, the choice is about content and your background, not logistics. The one logistical difference is language: if you prefer to test in a language other than English, AI-200 is offered in more of them, while Microsoft’s AI-300 page lists English only.
What each exam covers
AI-200 has four domains:
- Containerized solutions (20-25%): Container Registry, App Service, Container Apps with KEDA, AKS.
- Data management for AI (30-35%): Cosmos DB for NoSQL, Azure Database for PostgreSQL, Azure Managed Redis, with vector search in all three.
- Connecting to services (20-25%): Service Bus, Event Grid, Functions.
- Security, monitoring, troubleshooting (20-25%): Key Vault, App Configuration, OpenTelemetry, KQL.
AI-300 has five domains:
- MLOps infrastructure (15-20%): Azure Machine Learning workspaces, compute, assets, and infrastructure as code.
- Model lifecycle and operations (25-30%): training orchestration, registration, endpoints, safe rollout, drift.
- GenAIOps infrastructure (20-25%): Microsoft Foundry resources, model deployment, prompt versioning.
- Quality assurance and observability (10-15%): evaluation, safety checks, monitoring, tracing.
- Optimizing generative AI systems (10-15%): tuning retrieval-augmented generation and fine-tuning.
Weightings are as shown on the study guides we read; confirm them on the pages before you book. The AI-200 guide itself shows a small inconsistency for data management, so look at the live version.
Where they overlap
The overlap is real but thin.
- Both assume Python and a comfort with the Azure CLI or similar tools.
- Both care about identity: managed identities and role-based access appear in each.
- Both include observability. AI-200 asks for OpenTelemetry tracing and KQL; AI-300 asks for tracing, logging, and monitoring of generative AI apps.
- Both touch retrieval for AI. AI-200 looks at it as storage and search in databases; AI-300 looks at tuning the retrieval system for quality.
Do not expect studying one to cover the other. Microsoft lists no shared domain.
Where they differ
- Center of gravity. AI-200 is about the services an application calls. AI-300 is about the lifecycle of models and generative AI systems.
- Starting background. Microsoft’s AI-200 profile lists SDKs, data services, monitoring, messaging, vector databases, Python, and containers. The AI-300 profile asks for a data science background, Python, and an entry-level grasp of DevOps such as GitHub Actions.
- Infrastructure style. AI-300 names infrastructure as code with Bicep and the Azure CLI. AI-200 names Kubernetes manifests and Container Apps configuration.
- Evaluation. AI-300 includes quality and safety metrics for generative AI. AI-200 does not list evaluation.
How to choose
Ask three questions.
- What do you build or run today? Services and APIs point to AI-200. Models, pipelines, and prompts in production point to AI-300.
- Where are your gaps? If you have never deployed a container or used a queue, AI-200 fills gaps you will need anyway. If you have never registered a model or monitored drift, AI-300 does.
- What does your team need next year? Pick the exam that matches the work, not the one that sounds more advanced. Both are intermediate.
Quick guide
| You are… | Start with |
|---|---|
| A cloud or back-end developer adding AI features | AI-200 |
| A data scientist moving toward production | AI-300 |
| A DevOps or platform engineer supporting AI teams | Either; choose by whether you run apps (AI-200) or model workflows (AI-300) |
| New to Azure | A fundamentals exam first, then the one that fits your role |
Taking both
Taking both can make sense for a person who builds the application and also operates the model behind it. If you do, take them in the order your work needs. Study time does not add up neatly, because the content barely overlaps. Plan two separate schedules, and remember each credential renews on its own yearly clock; see renewal and retakes.
More guides for AI-200
- AI-200 Exam Format, Scoring and Retakes
- AI-200 Study Plan: 4-Week and 8-Week
- AI-200 Sample Questions With Answers
- AI-200 Prerequisites: What to Know First
- Is AI-200 Worth It? An Honest Look
- AI-200 Cheat Sheet: Key Facts to Review
- AI-103 vs AI-200: Which Exam Should You Take?
Next step
If AI-200 is your pick, begin with the AI-200 overview, then check the domains explained page. If AI-300 fits better, start at the AI-300 overview. For practice, see our AI-200 practice tests and AI-300 practice tests; they are original practice sets and complement Microsoft’s own learning material.
Frequently asked questions
Do I need AI-200 before AI-300?
Microsoft lists no prerequisite exam for either. Choose by your role and background.
Which one is harder?
It depends on you. AI-200 is broad across many services; AI-300 asks for both machine learning operations and generative AI operations. Read the AI-200 difficulty page and the AI-300 equivalent.
Are both credentials renewed the same way?
Yes. Microsoft says associate credentials expire annually and renew through a free online assessment.
Which has more weight on generative AI?
AI-300. Its three generative AI domains together are a large share of the exam. AI-200 uses AI as the setting but tests back-end services.