CCAR-F exam format
Last verified October 27, 2026. Source: Exam Guide v1.0, sections 3 to 6 and 9.
At a glance
| Items | 60 |
|---|---|
| Item types | Multiple-choice (one answer) and multiple-response (the item states how many to select, typically "Select TWO") |
| Structure | 4 scenarios presented, drawn at random from a bank of 6; each scenario frames a block of items |
| Time | 120 minutes, which is 2 minutes per item |
| Scoring | Scaled 100–1,000, pass at 720 |
| Delivery | Pearson VUE online proctored or test center, with an NDA accepted on screen before the first item |
How a form is built
Every candidate sees four of the six scenarios: Customer Support Resolution Agent, Code Generation with Claude Code, Multi-Agent Research System, Developer Productivity, Claude Code for Continuous Integration, and Structured Data Extraction. Within each block the items are written against the 30 task statements, weighted 27 / 18 / 20 / 20 / 15 across the five domains. A scenario does not restrict the domains it can touch: a support-agent block can ask about hooks, escalation, tool design and context management in consecutive items.
What the items look like
The 12 official samples in the guide are the style benchmark: a concrete production situation with numbers ("12% of cases skip get_customer"), four options that are all plausible to a beginner, and one that is best for the stated constraint. The stem endings recur: "most effective first step", "most likely root cause", "how should you restructure", "where should you create". Distractors fall into recognizable families: prompt-only fixes for rules that need enforcement, over-engineered classifiers and routing layers, fixes for a different problem, blame on a downstream component that is behaving correctly, invented flags and files, and popular misconceptions (a bigger context window fixes attention; sentiment predicts escalation).
What the samples get wrong
Candidates who passed are consistent: the real items are longer, hide the qualifier that decides the answer in the middle of the stem, and more often present two options that are both reasonable in general. The samples are also all single-answer, while the guide specifies multiple-response items too. Practice accordingly: our forms include select-two items and a hard tier where two options are defensible and only one stated detail separates them.
Timing
Two minutes per item is enough if you do not re-read scenarios. Each scenario's context is shown at the start of its block; read it once carefully, then rely on the stem. Flag anything you spend more than three minutes on and come back at the end. Candidates report using most of the 120 minutes including review, so do not plan to finish early.