If you are part-way through studying for the AWS Certified Machine Learning Engineer - Associate exam, you have a booking decision to make in the next few weeks. AWS is replacing MLA-C01 with MLA-C02, and the English version of the current exam stops being delivered on 28 September 2026.
The short answer: if you are already deep in MLA-C01 preparation and can get a seat before 28 September, sit MLA-C01. The certification you earn is identical either way, and MLA-C01 is a shorter exam on material you have already revised. If you are starting from scratch today, do not rush a half-prepared attempt. Take the MLA-C02 beta from 29 September at $75, or wait for general availability on 14 January 2027.
Here is what AWS has actually confirmed, and what changes between the two versions.
The MLA-C01 and MLA-C02 dates that matter
AWS published the transition timeline in its September 2026 certification update. These are the four dates worth putting in your calendar.
- 1 September 2026: registration opened for the MLA-C02 beta exam, in English only.
- 28 September 2026: the last day MLA-C01 is delivered in English.
- 29 September 2026: MLA-C02 beta delivery begins.
- 14 January 2027: MLA-C02 reaches general availability in all supported languages.
One detail catches people out. MLA-C01 does not disappear entirely on 28 September. It remains available in Japanese, Korean and Simplified Chinese throughout the beta period, right up to the 14 January 2027 general availability date. Only the English delivery stops in September.
Exam Tip: Booking closes before the exam does. Pearson VUE and PSI seats for a retiring exam thin out fast in the final fortnight, and online proctored slots go first. If you intend to sit MLA-C01, book the seat now and revise around the date rather than the other way round.
What changed in MLA-C02
The structural answer is: less than you might fear. AWS kept the same four content domains and added no new ones. What changed is the task statements and skills inside those domains, updated to match how the ML engineer role is actually practised now.
The additions cluster around generative AI:
- Generative AI implementation, including Amazon Bedrock and retrieval-augmented generation (RAG) architectures.
- Foundation models and large language models, covering selection, fine-tuning and operationalisation.
- Agentic AI, meaning the orchestration of AI agents and multi-step workflows.
- Responsible AI practices, applied across both traditional ML and generative AI.
That last point is the one candidates underestimate. Responsible AI is not confined to a governance question at the end of the paper. It runs through model selection, evaluation and monitoring, which means it can surface in any of the four domains.
If your MLA-C01 revision was built around SageMaker pipelines, feature engineering and endpoint autoscaling, that work still counts. The generative AI content sits on top of it rather than replacing it.
MLA-C01 vs MLA-C02 side by side
| MLA-C01 | MLA-C02 (beta) | |
|---|---|---|
| Questions | 65 (50 scored, 15 unscored) | 85 |
| Duration | 130 minutes | 170 minutes |
| Price | $150 USD | $75 USD during beta |
| Languages | English until 28 Sep 2026, then JA/KO/zh-CN until GA | English only during beta |
| Passing score | 720 on a 100 to 1,000 scale | Not published during beta |
| Content domains | Four | The same four |
| Results | Issued at the test centre | Delayed until AWS analyses beta data |
The domain weightings on MLA-C01, taken from the official AWS exam guide, are:
- Data Preparation for Machine Learning (ML) - 28% of scored content
- ML Model Development - 26%
- Deployment and Orchestration of ML Workflows - 22%
- ML Solution Monitoring, Maintenance, and Security - 24%
AWS has confirmed the domain structure carries over to MLA-C02. Treat the exact percentages as provisional until the MLA-C02 exam guide is published, and check the official guide before you build a revision plan around them.
Should you sit MLA-C01 before 28 September?
The honest answer depends on where your preparation already is, not on which exam looks better on paper. Both routes award the same credential.
Sit MLA-C01 if you are already prepared
You have done the SageMaker labs, you are scoring consistently on practice questions, and you can find a seat before 28 September. MLA-C01 is 65 questions in 130 minutes against 85 in 170. You get your result at the test centre instead of waiting on beta scoring. There is no advantage in throwing away revision that is already done.
Take the MLA-C02 beta if you want the discount and can handle the wait
At $75 the beta is half price, and you get 170 minutes for it. The trade is that beta results are not issued immediately, because AWS is using the sitting to evaluate question quality. If you need the certification on your CV by a fixed date, that delay is a real problem. If you do not, the beta is the cheapest route to the same certification.
Wait for general availability if you need certainty
From 14 January 2027, MLA-C02 is a normal exam: full language support, published exam guide, immediate results. Candidates who are starting from zero today, or who need study material in a language other than English, are usually better served waiting.
Exam Tip: A rushed attempt at a retiring exam is the worst of both options. You pay $150, you fail, and the version you revised for is gone. If you are not already scoring above the pass mark on practice questions, the 28 September deadline is not your deadline.
How to use the next few weeks if you are sitting MLA-C01
Twenty-four days is enough to finish preparation, not to start it. Spend it on the two domains that carry the most scored content and the ones candidates most often underestimate.
- Data Preparation, 28%. The largest single domain. Work through ingestion and transformation with Glue, feature engineering in SageMaker Feature Store, and data validation. Know when to reach for each service rather than memorising every parameter.
- Monitoring, Maintenance and Security, 24%. The second largest, and the one that trips up developers who came to ML from a modelling background. Model Monitor, data drift detection, IAM scoping for training jobs and endpoint encryption all appear here.
- Question formats. MLA-C01 uses ordering and matching questions alongside multiple choice and multiple response. Ordering questions give no partial credit: you place three to five items in the correct sequence or you score nothing. Practise those specifically.
- The compensatory model. You do not need to pass every domain, only the exam overall. That means a weak domain can be carried by a strong one, so put your remaining hours into raising your worst area rather than polishing your best.
For the full revision structure, our 8-week MLA-C01 study plan breaks this down week by week. If the generative AI content in MLA-C02 is new to you, the AWS AI Practitioner course covers Bedrock, foundation models and responsible AI at a level that transfers directly.
This is not the only AWS exam changing
MLA-C02 is one of three AWS exam updates announced in the same September 2026 cycle. AWS Certified Solutions Architect - Professional moves to SAP-C03 with SAP-C02 closing on 16 November 2026, and AWS Certified Developer - Associate moves to DVA-C03 with DVA-C02 closing on 30 November 2026. Both go to general availability the day after their predecessor closes.
If you hold or are studying for either of those, we have covered the SAP-C03 and DVA-C03 release dates separately. And if you came to MLA-C01 after the older Machine Learning Specialty was withdrawn, the MLS-C01 retirement post explains how that path folded into this one.
Frequently Asked Questions
Is the AWS Certified Machine Learning Engineer - Associate worth it?
It is the associate-level credential for engineers who build and operationalise ML pipelines on AWS, rather than data scientists who build models. It is worth it if your work involves SageMaker, deployment and monitoring. With MLA-C02 adding generative AI, foundation models and agentic workflows, it now covers a broader slice of what ML engineering jobs actually ask for.
How difficult is the AWS Certified Machine Learning Engineer - Associate?
AWS expects at least one year of hands-on SageMaker experience plus a year in a related role such as backend developer, DevOps engineer, data engineer or data scientist. Candidates without that background find it harder than the associate label suggests, mostly because of the deployment, CI/CD and monitoring content rather than the modelling.
What is the passing score for MLA-C01?
MLA-C01 is scored on a scaled range of 100 to 1,000, with a minimum passing score of 720. Scoring is compensatory, so you need to pass the exam overall rather than each domain individually. AWS has not published a passing score for the MLA-C02 beta.
Do I have to retake anything if I hold MLA-C01?
No. If you pass MLA-C01, your certification remains valid for its full three-year term. The exam version you sat is not shown on the credential, and there is no requirement to move to MLA-C02.
Can I still sit MLA-C01 after 28 September 2026?
Only in Japanese, Korean or Simplified Chinese. Those language versions run until MLA-C02 reaches general availability on 14 January 2027. English delivery of MLA-C01 ends on 28 September 2026.
Ready to Start Practising?
Whichever version you book, the thing that moves your score is answering exam-style questions under time pressure and reading the explanation for every one you get wrong. That is what turns service knowledge into exam performance, particularly on ordering and matching questions where partial credit does not exist.
Create a free CertCrush account and start practising against real exam-style questions with full explanations. If your target has shifted to the generative AI content in MLA-C02, start with the AWS Certified AI Practitioner course and build from there.