For AWS Certified AI Practitioner (AIF-C01), know how tokens represent model input and output, how embeddings represent information for similarity and retrieval, and how both fit into the foundation model (FM) lifecycle. The exam expects you to recognize these concepts and choose suitable approaches—not build tokenizers, train models, or implement AI/ML pipelines.
AWS’s 2026 exam guide assigns 24% of scored content to Fundamentals of GenAI and 28% to Applications of Foundation Models. Together, those domains account for 52% of scored content, calculated from the published weights. That is a useful study priority, not a promise that a particular exam form will contain a fixed number of questions on these topics. AWS AIF-C01 exam guide
As an Amazon Associate I earn from qualifying purchases.
What are tokens, embeddings, and vectors?
Tokens: units processed by a language model
A token is a piece of text a language model processes or generates. Depending on the tokenizer, it may correspond to a word, part of a word, punctuation, or another text unit; it is not reliably the same as a word or character. Token counts matter because AIF-C01 includes token-based pricing and its effects on inference cost and performance. The exact billing rules and price depend on the model and current terms, so do not assume a universal price per token.
Free tools Windows power users keep installed
One-click scans. No signup required.
Embeddings: numerical representations for comparison
An embedding is a numerical representation of information used to compare items by similarity. In a retrieval workflow, text or other content can be represented as embeddings so a system can find relevant material for a query. The exam-relevant distinction is that tokens are associated with model input and output, while embeddings support representation, similarity, and retrieval.
#1 Best Overall
Vectors: the form used in similarity workflows
A vector is an ordered set of numbers. Embeddings are commonly represented as vectors, which can be compared to identify content that is similar or relevant. AIF-C01 names tokens, chunking, embeddings, and vectors as foundational GenAI concepts; candidates need to understand their roles rather than implement the underlying algorithms.
How do tokens, chunking, and embeddings fit into RAG?
Retrieval-augmented generation (RAG) combines retrieval of relevant information with a foundation model’s response generation. At a conceptual level, content is divided into chunks, represented as embeddings, and stored for retrieval. A query can then be used to retrieve relevant content, which is supplied to the model as context for generating a response. Chunking determines the units of content available to retrieval; tokens relate to what the model processes and generates. Embeddings and vectors enable similarity-based retrieval.
AWS’s AIF-C01 objectives connect RAG to business applications and name Amazon Bedrock Knowledge Bases as an example. They also list Amazon OpenSearch Service, Amazon Aurora, Amazon Neptune, and Amazon RDS for PostgreSQL as examples of services used to store embeddings in vector databases. These are exam-scope examples, not a ranking or a guarantee of availability in every region. AWS notes that its in-scope service list is non-exhaustive and can change. AWS in-scope services
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteWhat are the stages of the foundation model lifecycle?
AWS identifies this lifecycle sequence: data selection, model selection, pre-training, fine-tuning, evaluation, deployment, and feedback. Treat it as a way to understand the work around an FM, not as a claim that every project follows identical steps or repeats them in the same way. AWS AIF-C01 exam guide
1. Data selection
Choose and govern information appropriate to the intended model or adaptation. The exam objective names this stage; it does not require candidates to design data engineering systems.
2. Model selection
Match a model to the task and its constraints. AWS highlights cost, modality, latency, multilingual capability, model size and complexity, customization needs, input and output length, and prompt caching. Consider these together: a model suited to one task may not meet another task’s language, response-time, or context requirements.
3. Pre-training
Pre-training is a lifecycle stage and one of the approaches named in the exam objectives. For AIF-C01, recognize where it fits; model-building and training mechanics are outside the intended foundational scope.
4. Fine-tuning
Fine-tuning adapts a model and is another lifecycle stage and customization approach. AWS’s detailed objectives also identify instruction tuning, domain adaptation, transfer learning, continuous pre-training, and data-preparation considerations. Know the concepts and tradeoffs rather than implementation procedures.
5. Evaluation
Evaluation asks whether the model’s results are suitable and meet business objectives. AWS includes human evaluation, benchmark datasets, and metrics such as ROUGE, BLEU, and BERTScore. A metric or benchmark is evidence to consider, not a substitute for judging whether the model serves its intended use.
6. Deployment
Deployment makes a selected model available for use. At this stage, inference choices connect to token counts, cost, and performance. AWS expects candidates to understand the general token-based pricing model and its implications; current model-specific prices are not established here and can change.
7. Feedback
Feedback completes the conceptual loop by informing future improvement. AWS names feedback as a lifecycle stage but does not prescribe a particular feedback system in the exam objective.
Recommended Free Tools
How should you compare customization approaches?
AWS identifies pre-training, fine-tuning, in-context learning, RAG, and model distillation as customization approaches with cost tradeoffs. For the exam, focus on what kind of change each represents and what need it may address, rather than assuming one approach is always best.
Best Value
| Approach | What to recognize | Decision consideration |
|---|---|---|
| Pre-training | A named lifecycle stage and customization approach. | Consider it as part of the model’s development; AWS identifies cost tradeoffs among customization approaches. |
| Fine-tuning | A lifecycle stage and an approach for adapting a model; related objectives include instruction tuning, domain adaptation, transfer learning, and continuous pre-training. | Consider whether model adaptation is needed and account for its cost tradeoffs. |
| In-context learning | A customization approach named in the objectives. | Consider it alongside other approaches for the task; the exam emphasizes tradeoffs, not implementation detail. |
| RAG | Retrieves relevant information to provide context for generation; AWS names Amazon Bedrock Knowledge Bases as an example. | Consider it when the application needs retrieved information; it involves retrieval and model inference. |
| Model distillation | A customization approach named in the objectives. | Compare it with other options in terms of the application’s needs and cost tradeoffs. |
What should guide model selection and inference decisions?
Use the task and constraints to narrow choices, then evaluate whether the model performs well enough for the intended business outcome. AWS’s selection criteria span capability, user experience, and cost:
- Task and modality: Check whether the model supports the type of input and output required.
- Language: Consider multilingual capability when the application serves users in multiple languages.
- Latency and performance: Match response-time expectations to the use case.
- Model size and complexity: Consider whether the model is appropriate for the task’s demands.
- Input and output length: Check that the model can handle the expected context and response lengths.
- Prompt caching: Include it among the model-selection considerations identified by AWS.
- Cost: Account for token-based inference expense using the current terms for the particular model; do not rely on a generic token price.
- Evaluation evidence: Use human evaluation, relevant benchmarks, and metrics alongside the business objective.
What does AIF-C01 expect you to know?
AWS describes the target candidate as having up to six months of exposure to AI/ML technologies on AWS and using, but not necessarily building, AI/ML solutions. The guide’s scope is foundational: recognize concepts, understand use cases, and choose suitable approaches. Coding models, implementing data engineering, hyperparameter tuning, and building or deploying AI/ML pipelines are outside the target role’s expected scope. AWS AIF-C01 exam guide
The exam guide is versioned and objectives may change. AWS’s revisions page lists version 1.0, published March 26, 2026, and version 1.1, published April 30, 2026; AWS says guides are periodically reviewed and changes are published approximately one month before they are reflected on an exam. Version 1.1 added objectives including token-based pricing and context engineering. Check the live guide when preparing. AWS exam guide revisions
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
