AWS MLA vs Google PMLE: Which Machine Learning Engineer Cert in 2026?
By Kyle Tice · Updated October 19, 2026 · Facts checked against official exam pages
Pick the AWS ML Engineer – Associate if you deploy and operate models on AWS and want a cheaper, longer-lasting credential, especially right now while the new MLA-C02 beta costs $75. Pick the Google Professional Machine Learning Engineer if you work on Google Cloud or want the harder, professional-level signal that ML hiring managers have respected for years. Either way, both exams changed in 2026, so make sure your study material matches the current version.
| MLA-C02 | PMLE | |
|---|---|---|
| Provider | AWS | Google Cloud |
| Code | MLA-C02 | PMLE |
| Level | associate | professional |
| Cost | $150 | $200 |
| Format | Proctored exam | Proctored exam |
| Questions | 85 (beta); 65 at GA | 50–60 |
| Duration | 170 min | 120 min |
| Validity | 3 yrs | 2 yrs |
| Prerequisites | None (1 yr SageMaker experience recommended) | None (3+ yrs industry experience recommended) |
| Study time | 60–100 h | 80–120 h |
| Difficulty | 3/5 | 4/5 |
| Employer demand | 5/5 | 4/5 |
| Atlas score | 94/100 | 83/100 |
| Status | beta | active |
What changed in 2026
Both exams moved in the same direction: less classic ML in isolation, more generative and agentic AI on each vendor’s platform.
AWS retired the ML Specialty on March 31, 2026, which left the ML Engineer – Associate as its main ML engineering exam. Then it released a new version. As of October 2026:
| MLA-C02 (beta) | MLA-C01 | |
|---|---|---|
| Questions | 85 | 65 |
| Time | 170 minutes | 130 minutes |
| Fee | $75 (beta) | $150 |
| English availability | From Sept 29, 2026 | Ended Sept 28, 2026 |
| Recommended experience | 1 year with SageMaker AI, Bedrock and other ML services | 1 year with SageMaker |
If you test in English today, you take C02. The extra items in a beta are unscored but you still have to sit through them, which is why the exam runs almost three hours. Beta results take about five business days.
Google revised the PMLE exam guide to reflect the move from Vertex AI to the Gemini Enterprise Agent Platform, updates to its data and analytics stack, and “a greater emphasis on Google Cloud native solutions.” Many popular courses still teach Vertex AI by name, so cross-check every resource against the new exam guide.
Content and domains
Neither vendor shows domain weights on its main page, so use the exam guides for detail. At a high level:
Google PMLE assesses six abilities:
- Architect low-code AI solutions
- Collaborate within and across teams to manage data and models
- Scale prototypes into ML models
- Serve and scale models
- Automate and orchestrate ML pipelines
- Monitor AI solutions
AWS MLA covers data preparation, model development, deployment and orchestration, and monitoring, maintenance and security of ML workloads. C02 adds generative AI, agentic AI and foundation model content, which is why AWS now names Bedrock alongside SageMaker AI in the recommended experience.
The practical difference: PMLE leans toward architecture and judgment, including when a low-code or pre-built option beats a custom model. MLA leans toward operating AWS services correctly, including IAM, networking, cost and the many SageMaker AI features.
Difficulty
We rate the PMLE 4 out of 5 and the MLA 3 out of 5.
- PMLE is a professional exam. Google recommends 3+ years of industry experience, with at least one year on Google Cloud. Questions are long scenarios where several answers are technically possible and you must pick the one Google considers best practice. You get 2 hours for 50–60 questions.
- MLA is an associate exam. It is more direct: given a requirement, which service or setting solves it. The C02 beta’s 85 questions in 170 minutes make stamina a real factor.
If you already passed AIF-C01, MLA is the natural next step. If you are new to cloud ML, start with MLA even if you will later go for PMLE.
Cost and validity
As of October 2026:
| AWS MLA | Google PMLE | |
|---|---|---|
| Fee (US) | $150 standard, $75 during C02 beta | $200 + tax |
| Validity | 3 years | 2 years |
| Renewal | Retake current version | Retake the standard exam |
| Cost per year | $50 (or $25 at beta price) | $100 |
The PMLE costs twice as much per year of validity. Google’s renewal FAQ says professional certifications last two years, and the PMLE is not one of the exams with a shorter renewal option, so you sit the full exam again. Google does give certified people a 50% discount code toward a later exam, which helps.
The AWS cert carries a hidden bonus: passing MLA also renews your AWS AI Practitioner if you hold it.
Employer recognition
We score MLA at 5 out of 5 for employer demand and PMLE at 4.
MLA’s score reflects AWS’s market share and the fact that it is now the main AWS ML engineering credential since the Specialty retired. Job posts that still list “AWS ML Specialty” will usually accept MLA as the current equivalent.
PMLE’s score is lower only because fewer companies run ML on Google Cloud. Where they do, the PMLE is one of the most respected cloud ML certifications, and its professional level tells employers more about seniority than an associate badge does. Data-heavy companies on BigQuery and teams using Gemini models are the best fit.
For senior ML roles, neither replaces a track record. Interviewers will ask you about a model you shipped, how you monitored it and what broke. See what employers want.
Study effort
- AWS MLA: 60–100 hours. AWS Skill Builder’s exam prep plan is free. Udemy courses from Stephane Maarek and Frank Kane, and Tutorials Dojo practice tests, are popular, but confirm they cover C02’s GenAI and agentic content before you buy.
- Google PMLE: 80–120 hours. The Google Cloud Skills Boost ML Engineer path is the official route, and there is an official study guide book. Hands-on labs matter: the pipeline and serving questions assume you have used the tools.
Our AWS and Google Cloud study guides break these down week by week.
Who should take both
- ML platform engineers at companies that run workloads on both clouds.
- Consultants who move between AWS and Google Cloud clients.
- Engineers changing cloud employers who want to prove they can transfer skills.
Take the one that matches your current job first. The ML concepts overlap heavily, so the second exam is mostly a platform study, but budget at least 40 hours for it.
Who should skip
If you build LLM applications rather than train and operate models, look at a GenAI developer exam instead, such as AWS AIP-C01. And if you are not yet working in ML, a portfolio project that trains, deploys and monitors a model will help you more than either badge.
FAQ
Is the AWS ML Specialty still available?
No. It retired on March 31, 2026. Existing holders stay certified until expiry. AWS points ML candidates to the ML Engineer – Associate and the Generative AI Developer – Professional.
Should I wait for MLA-C02 to leave beta?
Only if you want to know your result fast. The beta costs half price but has longer delivery and delayed scoring. For most people, the $75 saving is worth it.
Is the Google PMLE harder than the AWS MLA?
Yes. It is a professional-level exam with longer scenarios and a recommended 3+ years of experience. MLA is associate-level.
Do I need to code for these exams?
You will not write code during either exam, but both assume you have built and deployed models. Reading Python and SQL is expected.
Compare all ML options on /rankings/.