The short version
Take the Associate first if you are newer to machine learning on Databricks or want a gentler start. Take the Professional directly if you already build, test, and monitor ML systems on the platform every day. Databricks requires no prerequisite for either exam, so the order is your choice. Take both in sequence if your goal is the full ML track. The facts below come from the Databricks certification pages and exam guides, read on October 22, 2026.
Side by side
| Item | ML Associate | ML Professional |
|---|---|---|
| Level | Associate | Professional |
| Audience, in Databricks’ words | Basic machine learning tasks on Databricks | Advanced ML engineering at enterprise scale |
| Scored questions | 48 | 59 |
| Time | 90 minutes | 120 minutes |
| Unscored items | Up to 10, not marked | Up to 10, not marked |
| Sections | 4: 38%, 19%, 31%, 12% | 3: 44%, 44%, 12% |
| Languages | English, Japanese, Portuguese (Brazil), Korean | English |
| Delivery | Online proctored or test center | Online proctored or test center |
| Recommended experience | About 6 months hands-on | About 1 year hands-on |
| Prerequisite | None required | None required |
| Validity | 2 years | 2 years |
| Renewal | Take the current exam | Take the current exam |
| Retake | After 14 days | After 14 days |
| Guide date | March 1, 2025 in the guide’s update table | September 30, 2025 |
The exact passing score is not published for either exam.
What each exam covers
Associate: Databricks Machine Learning (MLOps basics, ML runtimes, AutoML, Unity Catalog feature tables, MLflow tracking and registry), data preparation (summary statistics, outliers, imputation, one-hot encoding, log scale), model development (algorithm choice, imbalance, pipelines, Hyperopt, cross-validation, metrics, bias and variance), and deployment (batch, real-time, streaming, custom model endpoints, traffic splits).
Professional: Model Development (SparkML pipelines, scaling and tuning with pandas Function APIs, Optuna and Ray, advanced MLflow such as nested runs and custom model objects, advanced Feature Store topics like point-in-time correctness and online tables), MLOps (lifecycle management in a deploy-code setup, unit and integration testing, environment architectures with Databricks Asset Bundles, automated retraining, drift detection and Lakehouse Monitoring), and Model Deployment (blue-green and canary strategies, rollout with Model Serving, custom model serving).
Overlap and differences
Overlap. Both use MLflow, feature stores, Unity Catalog, estimators and transformers, and model serving. If you study one, you reinforce the other.
Difference in depth. The Associate asks what a thing is and which basic option fits. The Professional asks you to design: which environment layout, which test at which stage, which monitor on which table type, which rollout for high traffic.
Difference in focus. The Associate spends much of its weight on platform basics and on data and model fundamentals (imputation, metrics, cross-validation). The Professional drops those and puts 44% on MLOps, including testing and monitoring, which the Associate hardly touches.
Difference in tools. Professional topics you will not meet on the Associate include Optuna, Ray, Databricks Asset Bundles, and Lakehouse Monitoring. Associate-only topics include AutoML, Hyperopt, and imputation and encoding choices.
How to choose
- Newer to Databricks ML, or six months of practice or less: Associate.
- Strong in ML theory, new to the platform: Associate, as a map of the platform.
- A year or more running ML pipelines, monitoring, and deployments on Databricks: go straight to the Professional if you wish. Databricks does not ask you to hold the Associate first.
- Short on time: the Associate is shorter (90 minutes against 120) and has a more compact outline.
- Want the credential for a role listing: check what the listing names, then choose.
Our view, which is a suggestion and not a vendor rule: if you are on the fence, the Associate costs less study time, and the topics you learn carry into the Professional. If you find the Associate outline easy and have production experience, the Professional is a better test of your skills.
One attempt, or both?
Both credentials last two years, and renewal means taking the current exam. If you take both, plan the dates so they do not expire in the same month. The renewal and retakes page explains the rules, and the same rules apply to the Professional exam.
More guides for DATABRICKS-ML-ASSOCIATE
- Databricks ML Associate Exam Format and Retakes
- Databricks ML Associate Study Plan
- Databricks ML Associate Domains Explained
- Databricks ML Associate Sample Questions
- Databricks ML Associate Prerequisites
- Is Databricks ML Associate Worth It?
- Databricks ML Associate Cheat Sheet
- GenAI vs ML Associate: Which Databricks Exam?
Next step
Read the overview of each exam: Associate and Professional. Then check yourself with the Associate practice tests or the Professional practice tests.
Frequently asked questions
Do I need the Associate before the Professional?
No. Databricks lists no prerequisite for the Professional exam, though it recommends more experience.
Which is harder?
The Professional, by design: more questions, more time, and design-level scenarios. See the difficulty page for the Associate.
Can I study for both at once?
You can share MLflow and feature store study, but the two outlines differ enough that separate plans work better.
Do the exams use the same languages?
No. The Associate is offered in four languages, the Professional in English only.