TL;DR: AWS is replacing the AWS Certified Machine Learning Engineer – Associate exam. MLA-C02 is in beta now (English only, $75), and general availability is January 14, 2027, when MLA-C01 retires. The last day to take MLA-C01 in English was September 28, 2026, so English speakers who want the current exam have already missed it. MLA-C01 remains available in Japanese, Korean and Simplified Chinese until retirement. The big change is content: MLA-C02 adds foundation model and generative AI work, agentic workflows and observability, with Amazon Bedrock now a primary focus alongside Amazon SageMaker AI. Domain weights shift only slightly (28/24/24/24 versus 28/26/22/24). This guide lays out the dates, the differences, and a decision for each situation.
If you were planning to sit MLA-C01 this autumn, the calendar has moved under you. Here is what is confirmed, what is not, and what to do next.
Quick Facts
| Item | MLA-C01 | MLA-C02 |
| Status today | Retiring. English delivery ended September 28, 2026 | Beta, English only |
| Key date | Retires January 14, 2027 (AWS announcement) | General availability January 14, 2027, all languages |
| Beta registration | Not applicable | Opened September 1, 2026. Beta delivery began September 29, 2026 |
| Beta format | Not applicable | 170 minutes, 85 questions, $75 USD |
| Standard format | 130 minutes, 65 questions, $150 | Exam guide lists 65 questions (50 scored and 15 unscored). Duration and price not stated in the guide text we read |
| Passing score | 720 out of 1,000 | 720 out of 1,000 |
| Scoring model | Compensatory (no per-section pass needed) | Compensatory |
| Certification validity | 3 years | 3 years |
| Primary services | Amazon SageMaker and AWS ML engineering services | Amazon SageMaker AI and Amazon Bedrock |
| Recommended experience | 1+ year with SageMaker and related AWS ML services | 1+ year with SageMaker AI, Amazon Bedrock and related services |
The Beta Window Decision Calendar
Most comparison posts list differences. This one starts with dates, because the dates decide your options.
| Date | What happens | What it means for you |
| September 1, 2026 | MLA-C02 beta registration opens | Beta seats became bookable |
| September 28, 2026 | Last day to take MLA-C01 in English | Already passed |
| September 29, 2026 | MLA-C02 beta delivery begins | The new exam is available now, English only |
| Today (early October 2026) | MLA-C01 only in Japanese, Korean, Simplified Chinese. MLA-C02 beta in English | Your language narrows your choices |
| January 14, 2027 | MLA-C02 reaches general availability in all languages. MLA-C01 retires | Last day of the old exam. From here, there is only MLA-C02 |
What to do, by situation
| Your situation | Best option | Why |
| English speaker, ready to test soon | MLA-C02 beta | It is the only version offered in English right now, and the beta price is $75 |
| English speaker, not ready until 2027 | Study for MLA-C02 and test at general availability | MLA-C01 is gone in English, so study for the new blueprint |
| Japanese, Korean or Simplified Chinese speaker, ready now | MLA-C01 | It is still delivered in those languages until January 14, 2027 |
| Same languages, ready after mid-January | MLA-C02 | MLA-C01 will be retired |
| Already holding MLA-C01 | Nothing is forced today | Your certification stays valid for its 3 year term |
| Studying from older MLA-C01 books or courses | Add Amazon Bedrock and foundation model topics | The new exam names Bedrock as a primary focus |
A note on the beta: this one is longer than the guide’s format (170 minutes and 85 questions versus 65 questions). We did not find a beta end date or a results timeline in the pages we read, so check AWS’s beta details before you book.
MLA-C01 vs MLA-C02, Side by Side
Domain weights
| Domain | MLA-C01 | MLA-C02 | Change |
| Data preparation | 28% (Data Preparation for ML) | 28% (Data Preparation for ML and AI) | None |
| Model development | 26% (ML Model Development) | 24% (ML Model and FM Development) | Down 2 points |
| Deployment and orchestration | 22% (Deployment and Orchestration of ML Workflows) | 24% (Deployment and Orchestration) | Up 2 points |
| Monitoring, maintenance and security | 24% | 24% (Operating, Monitoring, and Securing) | None |
The weights barely moved. The names did. “ML Model Development” became “ML Model and FM Development,” where FM stands for foundation model. “Data Preparation for ML” became “Data Preparation for ML and AI.” The content inside each domain is where the real change sits.
Scope and wording
| Topic | MLA-C01 | MLA-C02 |
| Services named as primary | SageMaker and AWS ML engineering services | SageMaker AI and Amazon Bedrock |
| Model types | Traditional ML | Traditional ML and foundation models |
| Workflows | ML pipelines and CI/CD | Adds agentic workflows and observability |
| Version handling | Not highlighted in our read | Model version management listed in the model development domain |
| Cost | Not highlighted in our read | Cost optimization named in the operating domain |
| Recommended experience | SageMaker plus a related role | SageMaker AI plus Amazon Bedrock plus a related role, with traditional ML and generative AI |
Out of scope, as listed in each guide
| MLA-C01 | MLA-C02 |
| Full end to end ML solution architecture | Designing and architecting full end to end AI and ML solutions |
| ML strategy guidance | Setting up best practices and guiding ML strategies |
| Multi-domain ML work such as NLP and computer vision | Working deeply in two or more ML domains |
| Model quantization analysis | Integration with a wide array of services or new tools |
The pattern is useful. This is still an engineer exam. It tests whether you can implement, deploy and maintain solutions, not whether you can design a whole platform or set strategy. Architecture-level design is listed as out of scope.
What Is New in MLA-C02
Based on the exam guide’s overview, these are the abilities MLA-C02 says it validates.
| Ability | What it means in practice |
| Work with traditional ML models and foundation models | You may need to choose between training your own model and using a foundation model |
| Ingest, transform, validate and prepare data | Data work stays the heaviest domain at 28% |
| Select modeling approaches, train, tune hyperparameters, manage versions | Covers the model lifecycle, including versions |
| Choose deployment infrastructure and configure auto scaling | Infrastructure and scaling decisions |
| Set up CI/CD pipelines for ML workflow orchestration | Automation and orchestration |
| Build agentic workflows and maintain observability | New wording that points to agent-based systems and their monitoring |
| Monitor models, data and infrastructure | Operations |
| Secure AI and ML systems through access controls and compliance | IAM, encryption and compliance |
AWS’s September 2026 certification announcement describes the same direction in its own words: generative AI, agentic AI and foundation model or LLM workloads alongside traditional ML engineering, including RAG architectures, AI agent orchestration and responsible AI practices. The announcement is a summary. The exam guide is the document that defines what is tested, so use the guide as your checklist.
Question Formats: A Possible Difference
| Question type | MLA-C01 guide | MLA-C02 guide (as we read it) |
| Multiple choice | Yes | Yes |
| Multiple response | Yes | Yes |
| Ordering | Yes | Not listed |
| Matching | Yes | Not listed |
The MLA-C01 guide lists four question types. The MLA-C02 guide text we read lists only multiple choice and multiple response. That could be a real change or an omission in the summary we received, so treat it as unconfirmed. Do not drop your practice on ordering or matching questions until you check the guide’s format section yourself.
Where the Numbers Disagree Inside AWS’s Own Material
This is a small but real conflict, and it is worth knowing before you book.
| Item | One AWS page says | Another AWS page says |
| MLA-C02 question count | 85 questions (certification page and announcement, beta) | 65 questions, 50 scored and 15 unscored (exam guide) |
| MLA-C02 duration | 170 minutes (beta) | Not stated in the guide text we read |
| MLA-C02 price | $75 (beta pricing) | Not stated in the guide text we read |
The simplest explanation is that the 85 question, 170 minute, $75 format applies to the beta, while the guide describes the general availability exam. We cannot confirm that from the pages we read. Expect the standard exam to follow the MLA-C01 pattern of 65 questions, but wait for AWS to state the final duration and price.
Who Should Take Which Exam, in Plain Terms
| Profile | Recommendation | Reason |
| Data scientist moving into deployment | MLA-C02 | The exam covers deployment, scaling and CI/CD, now with foundation models |
| Backend or DevOps developer moving into ML | MLA-C02 | Deployment, orchestration and monitoring map to your existing skills |
| Data engineer | MLA-C02 | Data preparation is 28% of the exam |
| Someone who wants the older, narrower traditional ML scope | MLA-C01, only if you are in a language it still supports and can test before January 14, 2027 | The window is closing |
| Someone building generative AI applications on AWS and wanting a developer focus | Consider the AWS Generative AI Developer exam instead | See our AIP-C01 exam guide |
If you are weighing this against a different cloud or a data platform path, our guides on the best cloud computing certifications, DP-100 vs AI-300 and Databricks ML Associate vs ML Professional show the alternatives.
How to Study for MLA-C02
- Start with the official exam guide. Print the four domains and mark each task as strong, weak or new.
- Spend the most time on data preparation. At 28%, it is the heaviest domain in both versions.
- Learn the Amazon Bedrock basics. The guide names Bedrock as a primary focus, and many older courses barely cover it.
- Keep your SageMaker AI skills sharp. SageMaker AI remains the other primary service.
- Practice deployment decisions. Know when to pick which infrastructure and how auto scaling behaves.
- Build one small pipeline end to end. Include a CI/CD step, a monitoring step and an access control step.
- Review security and compliance. IAM, encryption and data protection show up across domains.
- Take timed practice questions. Use MLA-C01 practice questions for the shared topics, and remember to add the new foundation model and agent topics on top.
If you studied for MLA-C01 already
| Keep | Add |
| Data preparation, feature handling and validation | Foundation model selection and use cases |
| Training, tuning and evaluation | Version management for models |
| Deployment, endpoints and auto scaling | Agentic workflow orchestration and observability |
| CI/CD for ML | Amazon Bedrock capabilities |
| Monitoring, security, IAM and encryption | Cost optimization in the operating domain |
Most of your MLA-C01 study carries over. The additions are concentrated in foundation models, Bedrock and agents.
Career Value
The certification validates that you can build, operationalize, deploy and maintain AI and ML solutions on AWS. It targets people in backend developer, DevOps, data engineer and data scientist roles.
We have not included salary figures in this post. We could not tie a pay number to a source that separates people with this specific certification from everyone else with a similar job title, and AWS’s own pages do not publish pay. If salary matters to your decision, our artificial intelligence salary and jobs guide covers the wider market.
Common Mistakes
| Mistake | Why it hurts |
| Booking MLA-C01 in English | English delivery ended September 28, 2026 |
| Assuming MLA-C02 is a small refresh because the weights barely moved | The content inside the domains changed, with Bedrock and foundation models added |
| Skipping Amazon Bedrock | The guide names it as a primary service |
| Expecting the beta to match the final exam length | The beta is 170 minutes and 85 questions, while the guide lists 65 |
| Dropping ordering and matching practice because one guide summary omits them | Unconfirmed, so check the guide |
| Waiting for the final exam when you are ready now | The beta is available at $75, if you accept the longer format |
| Studying only from pre-2026 material | It will miss the new foundation model and agent topics |
What We Don’t Know Yet
| Open question | Why it matters |
| The MLA-C02 beta end date | Decides how long you can take it at beta pricing |
| When beta results are released | Affects your planning if you need the credential by a deadline |
| Final MLA-C02 duration and price | The guide text we read does not state them |
| Whether ordering and matching questions appear on MLA-C02 | Affects how you practice |
| Whether the 65 question count in the guide is the final format | The beta uses 85 |
| Exact task-level changes from MLA-C01 | AWS provides a comparison document, but we could not read it |
| Whether any MLA-C01 holders get a renewal benefit | We found no statement on this |
FAQS
What is the AWS MLA-C02 exam?
It is the updated version of the AWS Certified Machine Learning Engineer – Associate exam. It covers building, deploying and maintaining AI and ML solutions on AWS, now including foundation models, agentic workflows and Amazon Bedrock.
When does MLA-C02 launch?
General availability is January 14, 2027 in all languages. A beta has been open in English since late September 2026.
When does MLA-C01 retire?
January 14, 2027, according to AWS’s September 2026 certification announcement.
Can I still take MLA-C01 in English?
No. The last day to take MLA-C01 in English was September 28, 2026. It remains available in Japanese, Korean and Simplified Chinese until retirement.
How much does the MLA-C02 beta cost?
$75 USD, English only. The beta runs 170 minutes with 85 questions.
Is the beta worth taking?
It is the only English option right now and it is cheaper than the standard price. The trade-off is a longer exam. We did not find a beta end date or results timeline, so check AWS before you book.
What are the MLA-C02 domains?
Data Preparation for ML and AI (28%), ML Model and FM Development (24%), Deployment and Orchestration (24%), and Operating, Monitoring, and Securing (24%).
How different are the weights from MLA-C01?
Only slightly. MLA-C01 was 28/26/22/24. Model development dropped two points and deployment gained two. The content inside the domains changed more than the numbers.
What is the passing score?
720 out of 1,000 for both MLA-C01 and MLA-C02, with compensatory scoring, so you do not need to pass each section separately.
Do I need Amazon Bedrock experience?
AWS recommends at least 1 year with Amazon SageMaker AI, Amazon Bedrock and other ML engineering services for MLA-C02. Bedrock is named as a primary focus.
Will my MLA-C01 certification still be valid?
We found no statement that retirement of the exam shortens the certification. Certifications are listed as valid for 3 years. Check your AWS certification account for your own expiry date.
Should I take MLA-C02 or the Generative AI Developer exam?
MLA-C02 is an ML engineering exam that now includes foundation models. The Generative AI Developer exam is aimed at building generative AI applications. Choose the job you want to do.
