If you are searching for the AWS Certified Machine Learning Specialty in 2026, you are searching for an exam that no longer exists. AWS retired MLS-C01 on 31 March 2026. You cannot book it, you cannot sit it, and no amount of study material with "2026 edition" on the cover will change that.
The problem is that almost every page still ranking for the AWS Certified Machine Learning Specialty was written before the retirement and never updated. Course listings still take your money. Study guides still sell. Search results still describe the exam in the present tense. So the single most common question about this certification right now is the one nobody on page one is answering: what replaces it?
Short answer: three certifications replaced it, not one. AWS split the old Specialty exam into a foundational credential, an associate credential and a professional credential. The one most people should sit is the AWS Certified Machine Learning Engineer Associate (MLA-C01), and that exam has its own deadline of 28 September 2026 before it is replaced by MLA-C02. If you were planning to start studying, the clock matters more than the syllabus.
This guide covers what happened, what replaced it, which of the three you should actually take, and what your existing MLS-C01 badge is still worth.
What happened to the AWS Certified Machine Learning Specialty?
The AWS Certified Machine Learning Specialty was AWS's flagship machine learning credential for roughly six years. It was a 180 minute, 65 question exam costing $300, with a scaled pass mark of 750 out of 1,000. It leaned hard into classical machine learning theory: feature engineering, algorithm selection, hyperparameter tuning, SageMaker training jobs and evaluation metrics.
That design aged badly. By 2024 the job it described had split in two. On one side sat data scientists doing modelling work. On the other sat machine learning engineers doing pipelines, deployment and operations. And running underneath both, generative AI arrived and made a large chunk of the classical modelling syllabus irrelevant to the day job of most people building AI on AWS.
AWS responded by retiring the single Specialty exam and building a proper ladder in its place. The last day to sit MLS-C01 was 31 March 2026.
Exam Tip: A retired exam is not a revoked certification. If you passed MLS-C01 before it retired, your credential stays active for the full three years from the date you earned it. You just cannot renew it, because there is nothing to renew it with.
What replaces the AWS Certified Machine Learning Specialty in 2026?
AWS points former MLS-C01 candidates at three separate credentials. Which one is right depends entirely on what you do for a living.
AWS Certified AI Practitioner (AIF-C01)
The foundational credential. It covers AI and machine learning fundamentals, generative AI concepts, foundation model applications, responsible AI and AI security and governance. It is aimed at people who need to speak AI credibly rather than build it: analysts, product managers, consultants, security staff and developers moving sideways into AI work.
AIF-C01 is 65 scored questions in 90 minutes, costs $100, and needs 700 out of 1,000 to pass. The five domains are weighted 20% fundamentals of AI and ML, 24% fundamentals of generative AI, 28% applications of foundation models, 14% guidelines for responsible AI, and 14% security, compliance and governance.
This is the natural landing spot for anyone who was considering MLS-C01 without a hands-on ML background. Our 4-week AIF-C01 study plan walks through the whole syllabus, and there is a full AWS AI Practitioner practice exam on CertCrush.
AWS Certified Machine Learning Engineer Associate (MLA-C01)
The direct successor for practitioners, and the one AWS itself steers MLS-C01 candidates towards. It is broader and far more practical than the exam it replaced. Instead of asking which algorithm suits a given dataset, it asks how you would ingest the data, version the features, orchestrate the training pipeline, deploy the endpoint, monitor drift and secure the whole thing.
MLA-C01 is 85 questions in total (65 scored, 20 unscored) across 170 minutes, costs $150, and needs 720 out of 1,000. Four domains, weighted as follows.
| MLA-C01 domain | Weight |
|---|---|
| Data preparation for machine learning | 28% |
| ML model development | 26% |
| Deployment and orchestration of ML workflows | 22% |
| ML solution monitoring, maintenance and security | 24% |
If you want a week by week route through it, we have an 8-week MLA-C01 study plan.
AWS Certified Generative AI Developer Professional (AIP-C01)
The professional tier, and the newest of the three. It targets engineers building production generative AI applications on Amazon Bedrock: foundation model integration, retrieval augmented generation, agent orchestration, evaluation and guardrails.
AIP-C01 costs $300, has 65 scored questions plus 10 unscored, and needs 750 out of 1,000. Its five domains are weighted 31% foundation model integration, data management and compliance, 26% implementation and integration, 20% AI safety, security and governance, 12% operational efficiency and optimisation, and 11% testing, validation and troubleshooting. We covered it in detail in AWS Certified Generative AI Developer (AIP-C01) Explained.
The three replacements side by side
| AIF-C01 | MLA-C01 | AIP-C01 | |
|---|---|---|---|
| Level | Foundational | Associate | Professional |
| Cost | $100 | $150 | $300 |
| Scored questions | 65 | 65 | 65 |
| Time | 90 minutes | 170 minutes | Standard professional length |
| Pass mark | 700/1000 | 720/1000 | 750/1000 |
| Focus | AI and GenAI literacy | ML pipelines and MLOps | Production GenAI on Bedrock |
| Best for | Non-builders and career changers | ML and data engineers | GenAI application developers |
MLA-C01 vs the old MLS-C01: what actually changed
People often assume the associate replacement is simply an easier version of the Specialty. It is not. It is a different exam testing a different job, and treating your old MLS-C01 study notes as a shortcut is the fastest way to fail it.
The Specialty was deep and narrow. Roughly 36% of its scored content sat in a single modelling domain, with heavy emphasis on choosing between algorithms, reading confusion matrices and tuning hyperparameters. Exploratory data analysis took another 24%.
MLA-C01 flattens that. Modelling drops to 26%. Data preparation rises to 28%. Deployment and monitoring together account for 46% of the exam, which is more than the entire modelling and data analysis half of the old syllabus. It also brings SageMaker JumpStart, Amazon Bedrock and MLOps tooling into scope, none of which the original Specialty covered meaningfully.
In practice that means three shifts.
- Less maths, more infrastructure. You need to know how to get a model into production and keep it there, not derive why gradient boosting beat a random forest.
- Generative AI is now in scope. Foundation models appear in the associate exam. They never appeared in MLS-C01.
- Security and governance carry real weight. IAM boundaries, VPC endpoints, encryption of training data and model artefacts, and drift monitoring all show up in the 24% monitoring, maintenance and security domain.
Exam Tip: If you were part way through MLS-C01 preparation when it retired, keep your SageMaker knowledge and bin your algorithm theory revision. SageMaker is still the spine of MLA-C01. Classical algorithm selection is a much smaller slice of it.
The deadline nobody is talking about: MLA-C01 closes on 28 September 2026
Here is the part that most replacement guides miss. MLA-C01 is itself being updated. AWS has announced MLA-C02, and the changeover is weeks away, not months.
- 1 September 2026: beta registration opens for MLA-C02, in English only, through Pearson VUE.
- 28 September 2026: the last day to sit MLA-C01 in English.
- Early 2027: the standard, all-languages version of MLA-C02 is expected.
The MLA-C02 beta runs 170 minutes with 85 questions and costs $75, which is half the price of the current exam. AWS's stated reason for the update is that machine learning engineers now implement generative AI solutions, work with foundation models and large language models, and orchestrate agentic AI workflows. So MLA-C02 expands into fine tuning, retrieval augmented generation, Amazon Bedrock and responsible AI, on top of the traditional ML engineering content.
That creates a genuine fork in the road right now.
Sit MLA-C01 before 28 September if: you have already started studying, your study materials are current, you want a stable syllabus with a published exam guide, and you want the certification on your CV this quarter. Every practice question and course in existence targets C01 today.
Wait for MLA-C02 if: you have not started yet, seven weeks is not enough runway for you, your actual job involves generative AI and agentic workflows, and you are happy to sit a beta. The $75 beta price is a real saving, but beta results are not released immediately and the full exam guide only publishes when registration opens on 1 September.
Do not start a leisurely twelve week MLA-C01 plan in mid August. You will run into the deadline with two weeks of content left.
Which AWS machine learning certification should you take now?
Match the credential to where you actually sit, not to where the retired Specialty used to.
You are new to AI and want a credential that opens doors. Take AIF-C01. It is $100, it is the cheapest way to get an AWS AI badge, and it is the only one of the three with no meaningful prerequisites. Pair it with hands on work rather than treating it as a finish line.
You are a data engineer, DevOps engineer or backend developer moving into ML. Take MLA-C01, and take it before 28 September if you can commit to eight focused weeks starting now. This is the credential recruiters are actually listing.
You are already building generative AI applications on Bedrock. Go straight to AIP-C01. It is the only one of the three that tests production GenAI engineering at a professional depth, and it is a stronger signal than an associate badge you could pass on theory alone.
You already hold MLS-C01 and it is still in date. Do not rush. Use the time to sit MLA-C02 when it stabilises in 2027, or move up to AIP-C01 now if your work has shifted to generative AI. Recertifying into the associate tier from a specialty credential is a sideways move on paper, though it does keep you current in the eyes of an applicant tracking system.
You are chasing salary rather than a specific role. The professional tier pays best, but it is the hardest to pass without production experience. Most people get more return from AIF-C01 plus a real project than from a professional exam they scrape through.
Does your existing MLS-C01 certification still count?
Yes, with caveats.
Your certification remains valid for three years from the date you earned it. If you passed in March 2026, just before the cutoff, you are certified until March 2029. AWS honours the badge, it appears in your certification account, and it still qualifies you for AWS Certified benefits during that window.
What you lose is the renewal path. There is no MLS-C02. When your three years expire, the credential lapses and you will need to sit one of the replacements to stay current. That is worth planning for now rather than discovering in 2029.
There is also a slower problem: recruiter recognition. Job adverts have already begun switching from "AWS Certified Machine Learning Specialty" to "AWS Certified Machine Learning Engineer Associate". A credential nobody is screening for stops doing useful work long before it technically expires.
This is the same pattern we saw with the AWS Advanced Networking Specialty retirement. AWS is steadily collapsing its specialty tier into the associate and professional tracks, and the specialty badges lose market visibility faster than they lose validity.
What AWS machine learning certifications pay
Salary data for machine learning credentials is unusually noisy, because it mixes certification-holder surveys with job title averages. Two figures are worth knowing.
Skillsoft's 2026 IT Skills and Salary Report puts average earnings for AWS Certified Machine Learning Specialty holders at $171,725. That is a survey of certified professionals, which skews high because the people who sat a $300 specialty exam tended to already be senior.
ZipRecruiter's July 2026 data puts the average AWS Machine Learning Engineer salary in the United States at $128,769, based on job postings rather than certification holders.
The honest read is that the certification is a filter, not a multiplier. It gets your CV past screening for roles that already pay well. It does not add a fixed premium to whatever you currently earn. The engineers at the top of those ranges got there through shipped production systems, and the badge opened the door to the interview.
Frequently Asked Questions
What will replace the AWS Certified Machine Learning Specialty?
Three certifications replaced it. AWS Certified AI Practitioner (AIF-C01) covers foundational AI and generative AI concepts. AWS Certified Machine Learning Engineer Associate (MLA-C01) is the direct practitioner successor and covers ML pipelines, deployment and MLOps. AWS Certified Generative AI Developer Professional (AIP-C01) covers production generative AI development on Amazon Bedrock. Most former MLS-C01 candidates should take MLA-C01.
What was AWS Certified Machine Learning Specialty?
It was AWS's specialty level machine learning certification, exam code MLS-C01, retired on 31 March 2026. It ran 180 minutes with 65 questions, cost $300, and required 750 out of 1,000 to pass. Its four domains covered data engineering, exploratory data analysis, modelling and machine learning implementation and operations, with modelling accounting for the largest share at 36%.
Which AWS certification is best for machine learning?
For most people it is the AWS Certified Machine Learning Engineer Associate (MLA-C01), because it maps to the ML engineering role employers are hiring for and it sits at an accessible associate price of $150. Choose AIF-C01 instead if you are new to AI and need a foundational credential, or AIP-C01 if you build generative AI applications in production.
How much does an AWS Certified Machine Learning Specialty holder earn?
Skillsoft's 2026 IT Skills and Salary Report puts the average at $171,725 for MLS-C01 holders. That figure reflects a senior, self selecting population rather than a certification premium. ZipRecruiter's broader July 2026 average for AWS Machine Learning Engineer roles is $128,769.
Can I still book the AWS Machine Learning Specialty exam?
No. The last day to sit MLS-C01 was 31 March 2026 and AWS has removed it from scheduling. Any site still selling MLS-C01 exam preparation as current material is selling you a syllabus for an exam you cannot book. Check the exam code on anything you buy: if it says MLS-C01, it is out of date.
Ready to Start Practising?
If you are heading for MLA-C01 before the 28 September deadline, or starting with AIF-C01 as your first AWS AI credential, the fastest way to find your weak domains is to sit realistic questions early rather than after you have finished reading.
CertCrush has a full AWS Certified AI Practitioner practice exam with domain level scoring, so you can see exactly which of the five AIF-C01 domains is costing you marks before you book anything. Every question comes with a written explanation of why the correct answer is correct and why the distractors are not, which is where most of the learning actually happens.
Create a free CertCrush account and start practising today. Then come back and read our 8-week MLA-C01 study plan to map out the seven weeks you have left.