Get Your CertAIF-C01PlayCompete
‹ All AIF-C01 questions

AIF-C01 Practice Question 54: Top p Nucleus Sampling

AWS Certified AI Practitioner (AIF-C01) · Test 6 · question 4 of 10

Fundamentals of GenAImedium

While tuning a model in the Amazon Bedrock playground, a developer raises Top P from 0.5 to 0.9 and leaves every other inference parameter at its default. In Amazon Bedrock, what is the effect of that change?

Select one answer
Show answer and explanation

Answer:

  • D. It widens the candidate pool to the top 90 percent of the probability distribution, so less likely tokens can be selected

AWS defines Top P as *the percentage of most-likely candidates that the model considers for the next token*. The model computes the cumulative probability distribution over possible next tokens and considers only the top P percent of it. A higher value increases the size of the pool and allows the model to consider less likely outputs; a lower value trims the pool to the safest tokens. Going from 0.5 to 0.9 therefore makes the output more varied (option D).

- A is wrong because response length is controlled by a separate family of parameters: maximum response length, penalties and stop sequences. Top P has no bearing on how long the answer is. - B is wrong because that describes near-greedy decoding, which you approach by driving *temperature* toward zero. Top P defines which tokens are eligible, not how often the top one wins. - C is wrong because a fixed *count* of candidates is Top K, not Top P. Top K of 50 means 50 tokens; Top P of 0.9 means however many tokens it takes to accumulate 90 percent of the probability mass, which may be two tokens or two hundred, depending on the prompt.

Reference: Influence response generation with inference parameters

Practice Test 6 ›

Study guides for AIF-C01

AI engineer roadmap 2026: skills, certs and the orderAI engineer roadmap for 2026: skills to learn in order, AI engineer vs ML engineer, which certifications fit each stage, and what to skip.
How AI Is Changing Software Jobs: Where to PrepareHow AI is changing software jobs: what history shows, the evidence on junior developers, what Klarna learned, new AI roles and five domains to prepare for.
How AWS Certification Exams Work: Scoring and RenewalHow AWS certification exams work: levels, booking, question types, unscored questions, scaled scores, pass marks, validity and how recertification works.
What Is Amazon Bedrock? Models, Pricing and GuardrailsWhat is Amazon Bedrock? How a request reaches a model, what happens to your data, token pricing, Guardrails, Knowledge Bases, AgentCore and the exam traps.
Google Cloud Generative AI Leader vs AWS AI PractitionerGoogle Cloud Generative AI Leader vs AWS AI Practitioner: format, cost, topics, free preparation and which certification to take first.
AWS certification cost in 2026, and free ways to prepareWhat every AWS certification costs in 2026 in USD and INR, retake and reschedule rules, the 50% voucher, and the free official ways to prepare.
AWS Generative AI Developer Professional (AIP-C01) guideWhat the AWS Certified Generative AI Developer – Professional (AIP-C01) exam covers - format, cost, the five domains, the services, and how to prepare.
Best AI certifications 2026: AWS, Google, Microsoft, ClaudeThe best AI certifications in 2026 from AWS, Google, Microsoft, Claude and Databricks, by level, with price, status and which one to take first.
Which AWS certification should you get? 2026 roadmapWhich AWS certification to get first, by role and experience, with each exam's status in October 2026, what is changing, and prices in USD and INR.
What Is RAG? How Retrieval-Augmented Generation WorksWhat RAG is and how it works: chunking, embeddings, vector search, hybrid search, reranking, GraphRAG and agentic RAG on AWS, Google and Azure.
Prompt engineering vs RAG vs fine-tuningWhen to use prompt engineering, retrieval-augmented generation, fine-tuning or continued pre-training, compared on cost, freshness, hallucination and privacy.