MLA-C02 Exam Guide: What Changed, Domains, Dates, Cost and a Study Plan [2026 Guide]

The AWS MLA-C02 exam is the updated version of AWS Certified Machine Learning Engineer – Associate, and it is the only version that English-language candidates can book today. MLA-C01 closed in English on September 28, 2026. The new AWS MLA-C02 exam keeps the same four domains, the same 720 pass mark and three-year validity. What changed is the content: generative AI now runs through the whole blueprint, with Amazon Bedrock, retrieval-augmented generation (RAG), agentic AI workflows, foundation model fine-tuning and responsible AI added across every domain. The exam runs as an English-only beta, under the code ME1-C02, until general availability on January 14, 2027. Our DVA-C03 exam guide covers the parallel Developer – Associate change.

AWS MLA-C02 exam at a glance

The AWS MLA-C02 exam is the updated Machine Learning Engineer – Associate test. It keeps four domains and a 720 pass mark, and it adds Amazon Bedrock, RAG, agents and responsible AI. The English beta (code ME1-C02) costs $75. General availability starts January 14, 2027.

• Certification: AWS Certified Machine Learning Engineer – Associate, level Associate
• Beta exam code: ME1-C02 (the certification code is MLA-C02; the beta books under ME1-C02)
• Standard format (from January 14, 2027): 65 questions, of which 50 are scored and 15 are unscored; multiple choice and multiple response only
• Beta format (running now): 85 questions in 170 minutes, English only
• Passing score: 720 on a scaled score of 100 to 1,000; scoring is compensatory, so you need the overall pass mark, not a pass in each domain
• Price: $75 for the beta. AWS has not published the standard exam price yet
• Validity: 3 years
• Beta results: typically available within 5 business days, because beta items need statistical evaluation

AWS MLA-C02 exam key dates

Date	What happens
September 1, 2026	Beta registration opens (English only); the MLA-C02 exam guide is released
September 28, 2026	Last day to take MLA-C01 in English. MLA-C01 stays available in Japanese, Korean and Simplified Chinese until general availability
September 29, 2026	MLA-C02 beta delivery begins
January 14, 2027	General availability in all languages; MLA-C01 retires in all languages

One honest caution: the AWS certification blog gives January 14, 2027 for general availability, while the exam page still lists the GA dates as TBD. Plan around January 14, 2027, and re-check the exam page before you book. We covered the original announcement in our AWS certification changes 2026 post.

What changed in the AWS MLA-C02 exam

The updated exam reflects the broadened scope of the ML engineer role. The domain structure remains the same and no new domains were added, but every domain was renamed, two weights moved, and the official exam guide lists 45 new skills out of 107 in total. The headline additions are:

• Generative AI implementation: Amazon Bedrock, RAG architectures, knowledge bases, embeddings and vector databases (OpenSearch Service, Amazon RDS with pgvector, Amazon S3)
• Agentic AI: orchestrating AI agents and complex workflows, deploying agents, agent version management, and monitoring agent performance and coordination failures
• Foundation models and LLMs: selecting an FM in Amazon Bedrock, fine-tuning and continued pre-training, prompt management, and evaluation with NLP metrics such as BLEU, ROUGE and BERTScore, plus LLM-as-a-judge frameworks
• Responsible AI: Amazon Bedrock Guardrails, masking and redaction of training data, bias checks across text, image and numeric data, and validation of AI training data integrity

The question types also narrowed: the guide lists multiple choice and multiple response only, and a multiple response question earns credit only when you select every correct option. A smaller practical change: the exam uses short names for some AWS services, with a mapping list available through the Help button during the exam.

Some MLA-C01 content was deleted, so do not waste study time on it: optimizing models for edge devices with SageMaker Neo, bring your own container (BYOC) with SageMaker, and model size reduction techniques were all removed from the blueprint.

AWS MLA-C02 exam domains and weights

Domain	Weight	What is new in MLA-C02
1. Data Preparation for ML and AI	28%	Vector databases, embedding models, document chunking and metadata for RAG, masking and redaction, data for FM fine-tuning and distillation, multimodal data
2. ML Model and Foundation Model (FM) Development	24%	FM selection in Bedrock, fine-tuning strategy, RAG architecture patterns, NLP evaluation metrics, human evaluation frameworks
3. Deployment and Orchestration of ML and AI Workflows	24%	Bedrock Knowledge Bases, agent deployment, prompt management, FM deployment options, RAG system configuration and refresh pipelines
4. Operating, Monitoring, and Securing ML and AI Solutions	24%	Bedrock Guardrails, agent monitoring, FM inference cost and token usage monitoring, Bedrock evaluations, credential types for FM access

Two weights moved against MLA-C01: Model Development drops from 26% to 24%, and Deployment and Orchestration rises from 22% to 24%. Data Preparation stays the largest domain at 28%. In question terms, the standard exam has about 14 scored questions from Domain 1 and about 12 from each other domain, so no domain is safe to skip.

Who should take the AWS MLA-C02 exam, and when

The exam guide describes the target candidate clearly: at least 1 year of experience using Amazon SageMaker AI, Amazon Bedrock and other AWS services for ML engineering, plus at least 1 year in a related role such as backend developer, DevOps developer, data engineer or data scientist, with experience in both traditional ML and generative AI. If your work is SageMaker only, the new content is a real study block, not a light refresh.

Take the beta now if you need the credential before January 2027, you already work with Bedrock and SageMaker, and you accept a longer sitting: 85 questions in 170 minutes, at half the usual Associate price, with results in about 5 business days instead of on the day.

Wait for general availability if you want the standard 65-question format, mature third-party courses and practice exams, or a non-English exam. Candidates testing in Japanese, Korean or Simplified Chinese can still take MLA-C01 until January 14, 2027.

Already hold MLA-C01? Your certification stays active for its full three-year validity regardless of the version change, and you recertify by passing the latest version before it expires. There is no pressure to upgrade early.

A six-week study plan for the AWS MLA-C02 exam

• Week 1 – Domain 1, data foundations: ingestion and storage choices, Glue and Data Wrangler transforms, Feature Store, data quality and bias checks
• Week 2 – Domain 1, AI data: embeddings, vector database choices, chunking strategies for RAG, masking and redaction, preparing fine-tuning datasets
• Week 3 – Domain 2: classic model selection and tuning in SageMaker, then FM selection in Bedrock, fine-tuning versus RAG decisions, evaluation metrics (accuracy and F1 beside BLEU, ROUGE, BERTScore and LLM-as-a-judge)
• Week 4 – Domain 3: endpoint options, Bedrock Knowledge Bases end to end, agent deployment, prompt management, CI/CD for ML and AI workloads
• Week 5 – Domain 4: model and agent monitoring, drift, Bedrock Guardrails, IAM and credential choices for FM access, token and inference cost control
• Week 6 – exam mode: the AWS Official Practice Question Set and the Official Pretest on Skill Builder, timed practice under the real question mix, and a final pass over your weakest domain

Build one small project while you study: a Bedrock knowledge base over a set of documents, with Guardrails attached and one agent that calls a tool. That single build touches all four domains. AWS also publishes a free Exam Prep Plan on Skill Builder, ending with the official practice question set and pretest.

Exam traps to avoid

• Booking the wrong code: during the beta, the exam appears under ME1-C02, not MLA-C02. Search for the beta code when you schedule.
• Trusting third-party GA prices and durations: AWS has not published the standard MLA-C02 price or duration yet, and the exam page lists GA dates as TBD. Any site that states them as fact is guessing.
• Studying SageMaker only: 45 of the 107 skills in the guide are new, and most of them are Bedrock, RAG, agents and evaluation. A C01-era course alone does not cover the new exam.
• Revising deleted topics: SageMaker Neo and BYOC are out of scope now. Time spent there is time taken from Guardrails and knowledge bases.
• Partial answers on multiple response: you must select every correct option to earn credit, and unanswered questions score as incorrect. There is no penalty for guessing, so never leave a question blank.
• Chasing a perfect domain: scoring is compensatory. A weak domain does not fail you if the overall score reaches 720, and 15 of the 65 standard questions are unscored trials you cannot identify.

Frequently asked questions

Is the AWS MLA-C02 exam available now?

Yes, as an English-only beta under the code ME1-C02, with delivery since September 29, 2026. The standard exam reaches general availability in all languages on January 14, 2027, per the AWS blog; the exam page still shows GA dates as TBD.

How much does the AWS MLA-C02 exam cost?

The beta costs $75. AWS has not published the standard exam fee yet. The MLA-C01 fee was $150, which is the standard Associate price, but treat the MLA-C02 GA price as unconfirmed until AWS lists it on the exam page.

What is the passing score for MLA-C02?

720 on a scaled score of 100 to 1,000. The exam uses a compensatory scoring model, so you pass on the overall score and do not need to pass each domain separately.

Can I still take MLA-C01?

Not in English: the last day was September 28, 2026. MLA-C01 remains available in Japanese, Korean and Simplified Chinese until general availability of MLA-C02 on January 14, 2027, when it retires in all languages.

I already hold MLA-C01. Do I need to retake the exam?

No. AWS confirms the certification remains active through its full validity period regardless of when you earned it. When the three years run out, you recertify by passing the latest version of the exam, which will be MLA-C02.

Sources

AWS Certified Machine Learning Engineer – Associate exam page, aws.amazon.com/certification/certified-machine-learning-engineer-associate/ (what is new, key dates, beta information, validity). Exam Guide (MLA-C02), docs.aws.amazon.com (domains, weights, question types, passing score, target candidate, additions and deletions). AWS Training and Certification blog, September 2026 certification updates (beta registration, MLA-C01 last day, GA January 14, 2027).