The AWS Certified Machine Learning – Specialty (MLS-C01) exam has retired. Its last exam date was March 31, 2026, so you can no longer register for it. If you earned it, your credential stays valid for 3 years from the date you passed — but anyone planning to take it now needs a different target.
The good news: AWS replaced one specialty exam with a better-structured AI/ML track. The catch is there's no single direct successor — which cert you should take depends on what you actually do. This guide breaks down the replacements and who each one is for.
Why MLS-C01 Retired
AWS frames the change as part of evolving its AI/ML certification portfolio. The practical reason most of the industry points to: MLS-C01 was built for an era of training custom models from scratch. It emphasized algorithm selection, model training, and tuning — deep ML theory.
That's no longer how most teams work. Increasingly, organizations operationalize ML in production (MLOps) and build with managed foundation models like those in Amazon Bedrock, rather than training models from zero. MLS-C01's blueprint had no Bedrock or generative-AI content, so AWS retired it in favor of certs that match how AI/ML is built today.
AWS wasn't alone: Microsoft retired its own data-scientist exam, DP-100, on June 1, 2026, and replaced it with the MLOps-focused AI-300 — both clouds converged on operations-over-modeling within a few months of each other.
If You Already Hold MLS-C01
Your certification stays active for 3 years from your pass date, so many holders are valid into 2027–2029. You don't lose anything — but you can't recertify in the same exam, so plan your next cert around your current role.
What Replaced It: The New AWS AI/ML Track
AWS points former MLS-C01 candidates to a portfolio of certs. Here's the accurate map:
| Certification | Code | Level | Cost | Who it's for |
|---|---|---|---|---|
| AI Practitioner | AIF-C01 | Foundational | $100 | Familiar with AI/ML but not building it — analysts, PMs, adjacent roles |
| Machine Learning Engineer – Associate | MLA-C02 | Associate | $75 beta (GA price TBA) | ML engineers deploying and operationalizing ML and foundation models on SageMaker AI and Bedrock |
| Data Engineer – Associate | DEA-C01 | Associate | $150 | Data-pipeline-heavy backgrounds feeding ML systems |
| Generative AI Developer – Professional | AIP-C01 | Professional | $300 | Building generative AI apps with foundation models, RAG, and agents |
Machine Learning Engineer – Associate (MLA-C02) — the closest replacement
Update (October 2026): MLA-C01 retired in English on September 28, 2026. Its successor, MLA-C02, is in beta now ($75, 85 questions, 170 minutes) and goes GA on January 14, 2027. It adds RAG, vector databases, Bedrock knowledge bases, and agents. See our MLA-C02 study guide for the full breakdown. The MLA-C01 details below are kept for comparison.
If you were going to take MLS-C01 to validate hands-on ML work, the ML Engineer Associate is your most direct path — with one important caveat about level.
- Level: Associate (a step below MLS-C01's Specialty tier)
- Cost: $150 USD
- Questions: 65 (50 scored + 15 unscored)
- Passing score: 720 / 1,000
- Duration: 130 minutes
- Focus: Implementing ML workloads in production and operationalizing them — deploying, monitoring, and maintaining models, primarily on Amazon SageMaker.
The key difference: MLS-C01 tested modeling (choosing and tuning algorithms); MLA-C01 tests MLOps (shipping and running models). It's not a one-to-one equivalent. For most working ML engineers, the MLOps focus is actually closer to the day job — but if your strength was deep modeling theory, know that MLA-C01 covers it more lightly.
AI Practitioner (AIF-C01) — for fundamentals
A foundational, $100 cert for people who work around AI/ML without building it — product managers, analysts, IT support, and engineers who want to calibrate their knowledge before going deeper. It's a low-risk entry point, not a replacement for hands-on ML validation.
Generative AI Developer – Professional (AIP-C01) — where AWS is investing
This is the top of the new AI track and the area MLS-C01 completely missed. It validates building production generative AI on AWS — foundation models, RAG, vector databases, and agents. The exam went GA in 2026 after its beta ended March 31, 2026 (the same day MLS-C01 retired), and the standard version was updated to include Amazon Bedrock AgentCore.
If your work has shifted toward generative AI, this is the credential that matches it. See our full AIP-C01 study guide for the exam breakdown.
Which One Should You Take?
Pick by What You Actually Do
Build and deploy ML models? → MLA-C02 (Associate; $75 during beta). Need fundamentals or in an adjacent role? → AIF-C01 (Foundational, $100). Build generative AI apps with foundation models? → AIP-C01 (Professional, $300). Data-pipeline heavy? → Data Engineer – Associate.
For most engineers who were eyeing MLS-C01, the decision is MLA-C02 if your work is classic ML/MLOps, or AIP-C01 if you've moved into generative AI. They're not competing — many engineers will end up holding both as the field splits into "operationalize ML" and "build with foundation models."
If you're mapping out a longer AWS path, our AWS certifications roadmap for 2026 shows how these fit alongside the architect and developer tracks, and the AI engineer career path covers where the roles are heading.
Final Thoughts
MLS-C01's retirement isn't a loss — it's AWS catching its certifications up to how AI/ML actually gets built in 2026. Pick the replacement that matches your role, not the one with the most prestige, and you'll get more out of the prep.
The fastest way to be ready is to build the systems these exams test. Explore our AI & Machine Learning courses for hands-on training in SageMaker, Bedrock, and production AI on AWS.