How Many Questions Claude Certification Exams Have: CCAR-F Guide

Knowing how many questions claude certification exams have is the first number every candidate should confirm before booking a sitting. For the Claude Certified Architect, Foundations (CCAR-F) exam, that figure is 60 items in 120 minutes. The Claude Certified Developer, Foundations (CCDV-F) track runs 53 items in the same window. Both exams share a common scoring scale of 100 to 1000 and a passing mark of 720.

How Many Questions Does the CCAR-F Exam Present?

The CCAR-F serves 60 items in a 120-minute proctored session, giving you an average of two minutes per question. Every item is scenario-based, testing practical architectural judgment rather than definition recall. Items come in two formats: single-answer multiple-choice and multiple-response, where the question explicitly states how many correct options to select.

Each sitting draws 4 scenarios at random from a bank of 6. Those scenarios provide the narrative context for clusters of questions, which means two candidates sitting on the same day may encounter different narrative settings even though both face 60 items weighted identically across the five domains.

What Item Formats Appear on the Exam?

Two item formats appear throughout the CCAR-F:

• Multiple-choice: select exactly one correct response from four or five options.
• Multiple-response: select the number of correct responses stated in the question, for example "select two."

Because every multiple-response item declares how many answers are required, there is no ambiguity about whether a question wants one answer or several. In practice, multiple-response items are where most candidates leave points behind. A scenario might describe a broken agentic architecture pipeline and ask you to select the two changes that resolve the problem without introducing unnecessary complexity. Eliminating partially-correct distractors requires internalised design knowledge, not just surface familiarity with the concepts.

The exam's item design is deliberately practical. Per Anthropic's exam guide, the exam is scenario-based and tests judgment rather than recall. A question will rarely ask you to name a concept. It will ask you to apply it, diagnose a failure, or choose between two architecturally plausible but meaningfully different approaches.

What Is the Passing Score for Claude Certification?

The passing score is 720 on a 100-to-1000 scale. Your score report provides three pieces of information: a pass or fail result, your scaled score, and a percent-correct breakdown by domain. The domain-level breakdown is where most retake candidates find the most actionable signal.

Anthropic does not publish the raw-to-scaled conversion formula. On a rough linear reading, 720 out of 1000 implies approximately 41 to 42 of the 60 questions answered correctly, but the actual conversion accounts for item difficulty across scenarios, so you cannot use that estimate as a guaranteed raw target. Aim for broad mastery across all five domains rather than optimising for a minimum threshold.

The scaled scoring model means that a sitting which draws harder scenarios from the bank of 6 does not automatically disadvantage candidates. The scaling compensates for form difficulty, so your result reflects your preparedness rather than the luck of which 4 of 6 scenarios were selected.

How Are the Five Domains Weighted Across the 60 Questions?

The five domains are not weighted equally. Domain 1 carries more than a quarter of the exam on its own. The table below shows all five domains and their official weights per Anthropic's exam guide.

Agentic Architecture & Orchestration at 27% represents approximately 16 of the 60 items. The domain covers hub-and-spoke multi-agent designs, coordinator responsibilities, subagent context isolation, hook-based enforcement patterns, session management strategies, and decomposition approaches for complex tasks. The breadth of Domain 1 is the primary reason it dominates preparation discussions.

Domains 3 and 4 are tied at 20% each. Claude Code configuration and prompt engineering & structured output together account for around 24 of the 60 items, roughly 40% of the exam. Candidates who over-index on agentic architecture and under-prepare these two domains often find their score report showing weak domain-level performance exactly where they assumed prior experience would carry them.

Domain 5, context management & reliability, represents the smallest slice at 15%, but its concepts are embedded in scenarios across every other domain. Attention dilution in long agentic loops, stale context in resumed sessions, and summary-injection patterns for fresh agent starts all appear as constraints within Domain 1 scenarios. Treating Domain 5 as low-priority because it carries the lowest weight is a consistent prep error.

How Does the CCDV-F Developer Track Compare on Question Count?

The Claude Certified Developer, Foundations (CCDV-F) exam runs 53 items in 120 minutes at the same $125 price point. The table below shows the key format differences between the two active Foundations-level tracks.

The CCDV-F does not use a scenario bank. Items are written directly against the skills in each of its eight domains, so every sitting presents the same structural format. The CCDV-F's largest domain, Domain 2 (Applications and Integration), carries 33.1% of the exam, which is more concentrated than any single CCAR-F domain. Candidates preparing for both tracks need separate study plans; the domain structures share concepts but not weightings, item counts, or formats.

AI Skill Certs is an independent prep platform and is not affiliated with, endorsed by, or approved by Anthropic. The platform's practice exams mirror the real formats: 60 questions scored 100 to 1000 for CCAR-F, and 53 questions for CCDV-F.

How Is the Exam Delivered?

The CCAR-F is available online-proctored or at a Pearson VUE test centre. Both delivery modes present the same 60-item, 120-minute experience. The exam is closed-book: no notes, no access to documentation, and no external tools during the sitting.

That closed-book constraint has a direct implication for preparation strategy. The exam rewards pattern recognition and internalised judgment, not the ability to look up a configuration option. If your preparation relies heavily on having the Anthropic documentation open in a second window, you are building a recall dependency that will not be available when it matters most.

Online-proctored delivery requires a webcam and screen-sharing session with a proctor. Candidates must clear their workspace, close all applications, and complete an identity verification step before the exam begins. The closed-book format means the exam is not testing your ability to use the API in real time. You will be asked to read a scenario, evaluate options, and select the answer that reflects the best architectural judgment given the stated constraints. The goal of preparation is to internalise principles, not to rehearse syntax.

What Does the Scenario Bank Structure Mean for Preparation?

The scenario-bank mechanism has a direct consequence for candidates who may need to retake the exam. Because each sitting draws 4 scenarios from a bank of 6, a second attempt may surface two or more scenarios absent from the first sitting. Reviewing the questions you got wrong after a failed attempt is necessary but not sufficient. You must cover the full task-statement map.

Anthropic has not published the identities of the six scenarios, but the exam's 30 task statements across the five domains define the complete space of possible scenario content. Tool Design & MCP Integration concepts, for example, frequently appear not as isolated Domain 2 scenarios but as components embedded in Domain 1 agentic architecture narratives. A scenario describing a malfunctioning multi-agent pipeline will often require you to diagnose a tool-call routing failure alongside a fix to the coordinator logic, drawing simultaneously on Domain 1 and Domain 2 knowledge. The exam tests whether candidates can reason about Claude systems as integrated architectures, not as a collection of independent domain topics.

How Should Domain Weights Guide Your Study Allocation?

The simplest allocation heuristic is to match study time to domain weight. For a 30-hour preparation plan, that looks like the table below.

This allocation assumes a level starting baseline across all five domains. If you build multi-agent Claude applications daily, Domain 1 intuitions may already be solid, and you can redirect 2 or 3 hours toward Domain 3 or Domain 5, where textbook knowledge is more likely to substitute for hands-on experience.

The concept library at /concepts maps 174 atomic concepts to the 30 CCAR-F task statements. Drilling by task statement rather than by domain chapter trains you to recognise the specific scenario pattern each task statement describes, which is the skill that converts knowledge into correct answers on a closed-book sitting.

The CCAR-F costs $125 per attempt, and the credential is valid for 12 months from the date it is awarded. As of 3 June 2026, more than 10,000 individuals had earned a certification within the Claude Partner Network, which has drawn more than 40,000 applicant partner firms. At that scale, a first-attempt pass carries meaningful credential value beyond the technical knowledge it signals.