Domain 3 — Product and Model Selection (12%)
12% of the exam, roughly 7 questions. Two separate skills bundled together: picking the right feature, and picking the right model.
3.1 Picking the right feature
| Feature | What it is | Reach for it when |
|---|---|---|
| A plain chat | A single conversation, no setup | The task is one-off and self-contained |
| Projects | A workspace with its own instructions, its own uploaded knowledge, and its own chat history | The work recurs, or several conversations share the same background material |
| Artifacts | Substantial output rendered in its own panel beside the chat — documents, tables, pages, small tools — that you can iterate on and share by link | The output is something you will keep, revise or hand to someone, rather than read once |
| Research | Claude runs many searches and reads sources before answering, returning a cited report | The question needs current information from multiple sources and you need to see where each claim came from |
| Web search | Lighter-weight lookup during a normal conversation | You need one or two current facts, not a full report |
| Connectors | Links Claude to the systems you already use, so it can read from them directly | The material lives in a work system rather than in a file you can upload |
| File uploads | Attach documents, spreadsheets, images and PDFs to a conversation | The source material is a file — almost always better than pasting text and hoping |
| Incognito chat | A conversation that is not saved to your history | A sensitive one-off you do not want retained — note it is also never used for model improvement |
Frequent exam pairing — "recurring task, same background documents every time" means Projects. "Needs current information with sources" means Research. Confusing the two is a classic wrong answer.
3.2 Picking the right model
Claude comes in tiers that trade capability against speed and cost. The names and version numbers change; the shape of the trade-off does not.
| Tier | Character | Good for |
|---|---|---|
| Haiku | Fastest, cheapest | High volume, simple, repetitive work where speed matters more than depth |
| Sonnet | Balanced | The everyday default — drafting, summarising, analysis, most daily knowledge work |
| Opus | Most capable of the general tiers | Genuinely complex reasoning, long documents, work where quality justifies the extra time and cost |
| Fable | Highest capability tier | The hardest reasoning and long-running research or analysis |
The examinable judgement is proportionality. Using the most capable model for a task any tier handles wastes time and budget; using the fastest for genuinely hard analysis produces work you will have to redo. Start where the task sits, and move up only when you hit a real limit.
3.3 Context limits, in practical terms
A conversation has a finite working memory. Long conversations, large attachments and accumulated back-and-forth all consume it. Two symptoms tell you that you are near the edge:
- Claude starts losing track of instructions you gave earlier in the conversation.
- It refers vaguely to attached material rather than citing it precisely.
The remedies, in order: start a fresh conversation carrying forward a summary of what matters; move the persistent background into a Project so it does not have to be re-established each time; and attach only the documents actually relevant to the question rather than everything to hand.
3.4 Plans, in the terms the exam cares about
You are not asked to recite prices, which change. You are expected to know what capability tends to sit behind which kind of plan:
- Free — core chat and a limited number of Projects, with the tightest usage limits.
- Pro and Max — individual paid plans; higher usage limits, access to more capable models and heavier features, and expanded project knowledge capacity.
- Team and Enterprise — organisational plans that add the things a company needs: shared Projects with view or edit permissions, central administration, and organisational controls over what is enabled.
Usage on paid plans is governed by rolling windows rather than a simple daily count, and the specifics change over time. The examinable point is that limits exist, that heavier models consume them faster, and that choosing a proportionate model is therefore also a practical way to get more work done.
← Previous: Domain 2 — Output Evaluation and Validation | Next: Domain 4 — Workflow Integration and Solution Design →
Explore Other Certification Study Guides
Switch to another exam guide to prepare for your next Claude certification:
Claude Certified Architect — Foundations
Core architectural principles: agentic loops & orchestration, tool design & MCP, Claude Code workflows, prompt engineering, and context reliability.
Claude Certified Architect — Professional
Advanced enterprise architecture: solution design, system integration, evaluation frameworks, governance, safety & developer enablement.
Claude Certified Developer — Foundations
Build software on Claude: API integration, model selection, tools & MCP servers, agents vs workflows, security, safety & debugging.