AWS MLA-C02 beta replaces MLA-C01 exam ahead of January 2027 retirement

AWS is retiring its Certified Machine Learning Engineer – Associate exam, MLA-C01, and replacing it with a revised MLA-C02 version that puts generative and agentic AI at the centre of the syllabus. According to a guide published on dev.to, the MLA-C02 beta has been running since 29 September 2026, and candidates can no longer sit MLA-C01 in English.

Key dates

The dev.to guide lays out the timeline:

• 1 September 2026 — MLA-C02 beta registration opens (English only) and the exam guide is released.
• 28 September 2026 — last day to take MLA-C01 in English.
• 29 September 2026 — MLA-C02 beta delivery begins.
• 14 January 2027 — planned general availability in all languages, with MLA-C01 retired in all languages.

The 14 January 2027 date comes from AWS's September 2026 announcement, but the official exam page still lists general availability as "TBD". Until MLA-C01 is retired, it remains bookable in Japanese, Korean and Simplified Chinese.

Beta format and scoring

The beta runs 85 questions in 170 minutes at a discounted USD 75, in English only. Questions are multiple choice and multiple response; blank answers count as incorrect, while guessing carries no penalty. Delivery is through Pearson VUE, either at a test centre or online with a proctor.

The MLA-C02 exam guide describes 50 scored and 15 unscored questions, a scaled score between 100 and 1,000, and a minimum passing score of 720 — though that breakdown does not obviously match the 85-question beta, so treat the figures cautiously. The guide also notes that a pass/fail designation does not apply to the beta.

Under AWS's beta rules, a beta attempt is allowed only once, results arrive within five business days, and a pass earns the full certification, valid for three years. Recertification means passing the latest version of the exam. The duration and price of the standard MLA-C02 have not been announced; for comparison, MLA-C01 had 65 questions in 130 minutes at USD 150.

The four domains

AWS kept the four-domain structure but renamed the domains and adjusted two of the weights:

• Data Preparation for ML and AI — 28% (unchanged)
• ML Model and Foundation Model (FM) Development — 24% (down from 26%)
• Deployment and Orchestration of ML and AI Workflows — 24% (up from 22%)
• Operating, Monitoring, and Securing ML and AI Solutions — 24% (unchanged)

Most of the new material covers generative and agentic AI: vector databases, embeddings, getting documents ready for retrieval-augmented generation, choosing and fine-tuning foundation models in Amazon Bedrock, agents, and Amazon Bedrock Guardrails. AWS's comparison page also lists items dropped from the exam, including SageMaker Neo for edge devices and bring your own container with SageMaker.

Who should take it

The exam guide describes the target candidate as someone with at least one year of experience using Amazon SageMaker AI, Amazon Bedrock and other AWS services for ML engineering, plus at least a year in an adjacent role such as backend software developer, DevOps developer, data engineer or data scientist. Familiarity with both traditional ML and generative AI is expected. Designing complete end-to-end AI and ML solutions, and working deeply across two or more ML domains, are listed as out of scope.

The dev.to guide also suggests a roughly 50-hour study plan allocated by domain weight — 14 hours on data preparation and 12 hours on each of the other three domains — and advises anyone who prepared for MLA-C01 to put extra time into generative AI skills, which now show up in every domain.

Why it matters

The revision tracks where AWS sees machine learning work heading: the syllabus now weights Bedrock, foundation models, RAG and agents alongside traditional pipelines, while edge-deployment topics like SageMaker Neo fall away. For practitioners, the calculus is practical. English-speaking candidates can no longer sit MLA-C01, so the beta — at half the usual price — is the only route to the certification until general availability, currently pegged for 14 January 2027 but still marked TBD on AWS's exam page. Anyone mid-preparation for MLA-C01 needs to redirect toward the generative AI content, and the nearly even domain weights mean no section can safely be skimmed. As with any beta, the details are worth confirming on AWS's Machine Learning Engineer – Associate page before booking.