CCAR-P

By Anthropic · Browse Mode
About the Exam

Prepare for the CCAR-P certification from Anthropic with 10 practice questions, community-verified answers, and detailed explanations.

Exam Topics

  • Solution Design & Architecture17%
  • Claude Models, Prompting & Context Engineering13%
  • Integration19%
  • Evaluation, Testing & Optimization16%
  • Governance, Safety & Risk Management14%
  • Stakeholder Communication & Lifecycle Management14%
  • Developer Productivity & Operational Enablement7%

How to Use This Practice Exam

  1. Browse — Read each question, select your answer, and reveal the explanation.
  2. Exam Mode — Simulate real exam conditions with a timed session and score report.
  3. Learn Mode — Spaced repetition schedules questions you struggle with for long-term retention.

Download the Full Exam PDF

Get every question and answer in a clean, printable PDF built for offline study. Purchase once, keep permanent access, and re-download the latest version anytime.

Get PDF
Learn More

Last updated September 11, 2026 at 6:50 AM

Topic filter
Retired questions
Question sort
Questions per page

QuestionQ12

Evaluation, Testing & Optimization

You are evaluating model-selection claims used by a peer team.

For each of the following statements, select Yes if the statement is true. Otherwise, select No.

Yes or No
StatementsYesNo
Heavier-reasoning models typically consume more of the rate limit and run more slowly than lighter models on the same input.
A representative-sample evaluation on the target workload is the right basis for confirming a model selection.
The most capable model is always the right choice regardless of the workload profile.
Tiered routing can serve routine traffic with a lighter model while reserving a heavier model for traffic that needs its capability.
Selecting a model based on which version was most recently announced is the right basis for production decisions.
Explanation

Higher reasoning effort trades increased latency and token use for deeper reasoning, so it can consume more rate-limit capacity. Model selection should be confirmed through evaluations that represent the production workload and its quality, latency, and cost requirements. The most capable model is not invariably appropriate: smaller models can be better for cost-sensitive or latency-sensitive routine work. Tiered routing appropriately directs simple requests to a lighter model and reserves a stronger model for complex requests. Announcement recency does not establish production suitability; measured performance on the target workload does.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!

QuestionQ17

Solution Design & Architecture

You are classifying chunking strategies by the corpus type to which each is best suited.

For each chunking strategy, select the applicable corpus type: “Long Structured Documents,” “Heterogeneous Short Records,” or “Code or Hierarchical Specs.”

Select
function-level or section-level chunking for code modules
tree-aware chunking that follows code or specification hierarchy
per-record chunking where each record is one chunk
semantic chunking along clause or paragraph boundaries
fixed-size chunking with overlap for short records of similar length
hierarchical chunking that mirrors document section structure
Explanation

Function-level and tree-aware boundaries preserve the logical hierarchy of source code and specifications. Per-record and fixed-size overlapping chunks suit short, independently meaningful records. Semantic paragraph/clause boundaries and document-section hierarchies preserve the context and structure of long documents.

Community Discussion

No comments yet. Be the first to start the discussion!

QuestionQ20

Solution Design & Architecture

For each statement, select Yes if it reflects a sound architectural practice. Otherwise, select No.

Yes or No
StatementsYesNo
Authentication and authorization run before retrieval so retrieval can filter by identity.
An asynchronous queue absorbs bursty traffic between intake and the model invocation layer.
Tax computation is encoded directly in the system prompt rather than in code.
Conversation logs include unredacted government identifiers to maximize signal for tuning.
Explanation

Authorization before retrieval enables identity-based access filtering. Queues smooth burst traffic and decouple request intake from asynchronous model processing. Tax calculations require deterministic, testable application logic rather than prompt behavior. Government identifiers are sensitive PII, so logs should redact or mask them and restrict their use.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!

QuestionQ24

Solution Design & Architecture

You are ordering decomposed components in an invoice-processing pipeline.

For every decomposed component, choose the execution layer to which it belongs: “Pre-Processing”, “Model Stage”, or “Post-Processing”.

Select
Claude-based classification of invoice type
persistence of validated records to the system of record
schema validation of the extracted fields
Claude-based extraction of structured invoice fields
PII redaction on extracted text prior to model invocation
optical character recognition on scanned invoice images
Explanation

Pre-processing prepares invoice inputs for inference, including OCR and redacting sensitive text before model invocation. The model stage performs the Claude classification and structured-data extraction. Post-processing validates extracted output against its schema and persists only validated records to the system of record.

Community Discussion

No comments yet. Be the first to start the discussion!

That's the end of the preview

It's free

100% of the questions are free for all users.
No strings attached.

Sign up freeLog in
Topics covered
Solution Design & ArchitectureClaude Models, Prompting & Context EngineeringIntegrationEvaluation, Testing & OptimizationGovernance, Safety & Risk ManagementStakeholder Communication & Lifecycle ManagementDeveloper Productivity & Operational Enablement
Know a question that should be here? Contribute to this exam
Back home