OCI AI FOUNDATIONS
1Z0-1122 Study Guide: OCI AI Foundations, Topic by Topic
How to study for Oracle's OCI AI Foundations Associate exam, 1Z0-1122-26: what to know in each topic area, RAG or fine-tuning and other decisions, a worked defect-detector example and a six-week plan.
A 1Z0-1122 study guide should follow the six topic areas Oracle lists for its Oracle Cloud Infrastructure AI Foundations Associate exam, currently coded 1Z0-1122-26: AI foundations, machine learning, deep learning, generative AI and large language models, the OCI AI portfolio, and OCI Generative AI with Oracle Database AI Vector Search and Select AI. The exam is conceptual and multiple choice, so the work is learning definitions well enough to tell similar terms apart and matching a business need to the right Oracle Cloud Infrastructure (OCI) service. Study each area in order, work a few small calculations by hand, sit Oracle's practice exam on MyLearn, and book the real exam with Oracle, which sets any exam fee.
Checked in October 2026. Oracle's education site did not respond when we checked on 11 October 2026, so this guide does not quote the exam's question count, time, passing score or fee; read them on Oracle's own exam page before you book.
The exam code and what changed
Oracle refreshes its foundations exams roughly once a year and changes the suffix of the code. The current version, listed on Oracle MyLearn on 11 October 2026, is 1Z0-1122-26, which replaced 1Z0-1122-25. Study material written for the older version still covers most of the ground, but the generative AI part of the OCI portfolio moves quickly, so check every service name against Oracle's current documentation.
Oracle's OCI Generative AI documentation now describes the service in three parts: enterprise AI models (chat, embeddings, reranking and compatible APIs), enterprise AI agents that combine models with tools, memory and retrieval, and governance controls such as access policies, private endpoints and guardrails. Custom models can be imported, fine-tuned and hosted on dedicated AI clusters.
You will not write code. The exam asks whether you can explain AI, machine learning, deep learning and generative AI in plain terms, and whether you can pick the right OCI service for a described need. Basic school algebra is enough for the metric questions, and no cloud or programming experience is assumed.
Topic by topic: what to know cold
| Topic area | Know these terms | Typical trap |
|---|---|---|
| AI foundations | AI, ML, DL, generative AI; AI workloads | Calling every chatbot "machine learning" when it is generative AI |
| Machine learning | Supervised, unsupervised, reinforcement; regression, classification, clustering | Mixing up precision and recall |
| Deep learning | Neuron, activation, loss, backpropagation; CNN, RNN, LSTM, transformer | Using a CNN for a text sequence task |
| Generative AI and LLMs | Tokens, embeddings, attention, temperature, prompt, RAG, fine-tuning | Choosing fine-tuning for knowledge that changes monthly |
| OCI AI portfolio | OCI Language, Speech, Vision, Document Understanding, OCI Data Science, GPUs | Using Speech for text-to-text or Vision for invoices |
| OCI Generative AI and Vector Search | On-demand models, dedicated AI clusters, agents, AI Vector Search, Select AI | Thinking fine-tuning runs on shared on-demand capacity |
The decisions the exam keeps asking about
Many questions are a choice between two good-sounding options. Learn the pairs:
- Classification or detection? Classification labels a whole image; object detection finds and boxes each object in it.
- Transcription or text-to-speech? Speech-to-text turns audio into text; text-to-speech goes the other way.
- Precision or recall? Precision asks how many flagged items were right; recall asks how many real items were caught.
- RAG or fine-tuning? Retrieval-augmented generation adds knowledge at question time; fine-tuning changes the model's behaviour.
- On-demand or dedicated? On-demand calls pretrained models on shared capacity; a dedicated AI cluster reserves GPUs for fine-tuning and hosting your models.
| Need | Best first choice | Reason |
|---|---|---|
| Answers from documents that change monthly | RAG | Update the index, not the model; answers can cite sources |
| A consistent style, format or vocabulary | Fine-tuning | Behaviour is learned into the weights |
| A one-off task with clear instructions | Prompt engineering | Cheapest and fastest; try it first |
| Private data that must not train a shared model | RAG | Data stays in your store and is retrieved per query |
| Read fields from invoices | OCI Document Understanding | Pretrained key-value and table extraction |
| Search business data by meaning | AI Vector Search | Vectors stored next to relational data |
Oracle's AI Vector Search overview explains the idea behind the last row: vector embeddings, numbers that represent meaning, are stored in the database next to business data, so you can query by semantics rather than keywords.
Worked example: a defect detector's metrics
Metric questions on a foundations exam are short, but the fastest way to get them right is to have worked one example yourself.
Scenario. A vision model inspects 1,000 parts on a production line. In truth, 50 parts are defective. The model flags 60 parts, and 45 of those really are defective.
- True positives (TP) = 45; false positives (FP) = 60 - 45 = 15
- False negatives (FN) = 50 - 45 = 5
- True negatives (TN) = 1,000 - 45 - 15 - 5 = 935
- Accuracy = (45 + 935) / 1,000 = 0.98
- Precision = 45 / (45 + 15) = 0.75
- Recall = 45 / (45 + 5) = 0.90
- F1 = 2 x 0.75 x 0.90 / (0.75 + 0.90) = 1.35 / 1.65 = 0.818
Accuracy of 98 percent sounds superb, but it mostly reflects the 950 good parts. Recall of 0.90 says one defective part in ten still ships, and precision of 0.75 says one rejected part in four was fine. Which matters more depends on the cost of a missed defect against the cost of a false reject, and that is exactly the reasoning a scenario question tests.
Worked example: is your study plan long enough?
The free course below estimates 775 minutes of study across its modules. Suppose you can study three evenings a week for 45 minutes, for six weeks before your exam date.
- Time available each week: 3 x 45 = 135 minutes
- Over six weeks: 135 x 6 = 810 minutes
- Spare: 810 - 775 = 35 minutes
That is less than one evening. Add at least one more session for Oracle's practice exam and one to review the questions you got wrong, or move the exam back a week.
A six-week 1Z0-1122 study plan
- Week 1: enrol in Oracle's free learning path on MyLearn and note the current exam code and details. Study AI foundations and AI workloads.
- Week 2: machine learning: learning types, data splits, metrics, overfitting and the bias and variance trade-off.
- Week 3: deep learning: neurons, layers, CNNs, RNNs and LSTMs, and the transformer.
- Week 4: generative AI: tokens, embeddings, temperature and top-p, prompting, RAG and fine-tuning.
- Week 5: the OCI AI services and OCI Generative AI, AI Vector Search and Select AI. Write one line per service: input, output, one use case.
- Week 6: take Oracle's practice exam once, cold. Map every wrong answer to a topic, revise it, then book the real exam.
An OCI Free Tier account is useful for the optional console exercises. Oracle's Free Tier documentation describes Always Free resources that never expire plus a trial with credits valid for up to 30 days. OCI Generative AI is available only in some regions and is billed by use after the trial, so delete what you do not need.
What to avoid
Sites offering "real 1Z0-1122 questions" or "verified dumps" appear high in search results. Using them breaks Oracle's candidate rules and can cost you the credential, and their answers are often wrong. Practise on original questions and Oracle's own practice exam instead. For a wider view of free AI learning, see our list of the best free AI courses for engineers, and for how LLMs behave on technical work read generative AI training for engineers.
A free course built on this guide
The free Oracle Cloud AI Foundations Exam Prep (1Z0-1122) course follows all six topic areas in thirteen study modules, with lessons from IBM Technology, StatQuest, 3Blue1Brown, Cloud Alchemy Academy and Oracle's own channels, original notes, worked problems and original practice questions. Learning is free; the Oracle exam is booked and paid for separately with Oracle.
It is a free course with a verifiable certificate of completion, and anyone can check a certificate on our verification page. It is not the Oracle credential, which only Oracle awards after its own exam, and EDWartens is not an Oracle training partner; Oracle and OCI are Oracle's trademarks. Start the 1Z0-1122 exam prep course and work through it alongside Oracle's learning path.
Take the free course
Questions
What does the 1Z0-1122 exam cover?
Oracle lists six topic areas for 1Z0-1122-26: AI foundations, machine learning foundations, deep learning foundations, generative AI and large language models, the OCI AI portfolio, and OCI Generative AI with Oracle Database AI Vector Search and Select AI. Responsible AI runs through all of them.
What is the current code for the OCI AI Foundations exam?
1Z0-1122-26, as listed on Oracle MyLearn on 11 October 2026. It replaced 1Z0-1122-25, and Oracle changes the suffix roughly yearly, so check the code on Oracle's exam page before you book.
How long should I study for 1Z0-1122?
About four to six weeks at a few hours a week for a beginner. A free prep course of around 13 to 14 hours, Oracle's own learning path and one sitting of Oracle's practice exam with a review of every wrong answer is a realistic plan.
Do I need coding or cloud experience for OCI AI Foundations?
No. The exam is multiple choice and conceptual. Basic algebra helps with metrics such as precision and recall, and an OCI Free Tier account is useful but optional for trying the services.
Is the 1Z0-1122 exam free?
Do not assume so. Any exam fee is set by Oracle and paid to Oracle, and offers can change. Check the current fee, if any, on Oracle's own exam page before you register.
Is there a free 1Z0-1122 course with a certificate?
Yes. Oracle Cloud AI Foundations Exam Prep (1Z0-1122) on EDWartens is a free course with a verifiable certificate of completion, covering all six topic areas with original practice questions. It is not the Oracle credential, which only Oracle awards after its exam.
Sources
- Oracle Docs: Overview of Generative AI Service (read 11 October 2026)
- Oracle Docs: Overview of Oracle AI Vector Search (read 11 October 2026)
- Oracle Docs: Oracle Cloud Infrastructure Free Tier (read 11 October 2026)
Written by the EDWartens engineering team for general education. Product names are trademarks of their owners; mentioning them does not imply endorsement. Prices and terms of other providers were checked on the date shown and can change.
Free Generative AI and Prompt Engineering Courses With Certificate, Compared
Free Machine Learning Course for Mechanical and Electrical Engineers
Predict Machine Failure From Sensor Data in Google Colab (Video)
Machine Learning for Beginners: What It Is, How It Works and How to Start Free
Predictive Maintenance With AI and Machine Learning: How It Works
Natural Language Processing for Engineers: From Maintenance Logs to Manual Search