AI-900 Practice Question: Describe features of generative AI workloads on Azure

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  The visual display area where AI responses appear in a chat interface

  Why it's wrong here

  The visible panel where a chat interface renders responses is purely a client-side UI component; it does not constrain or define how much text the model can ingest. The context window is instead an internal, fixed token limit—typically thousands to millions of tokens—that governs prompt plus generated output. Confusing the display area with context capacity is like confusing a monitor with a computer's RAM.

• ✓

  The maximum amount of text an LLM can process in a single interaction

  Why this is correct

  A model's context window is the maximum number of tokens—roughly words or word fragments—that the model can accept as input and continue generating from in a single forward pass. This budget must be shared between the user prompt, including conversation history and any retrieved documents, and the model's response. If the total exceeds the context window, the model either raises an error or silently truncates earlier content, which is why a larger context window enables longer, more coherent multi-turn conversations.

• ✗

  The number of seconds before a model response times out

  Why it's wrong here

  A response timeout is an engineering safeguard that defines how long a client (like an application or SDK) will wait for an HTTP response before abandoning the call; it has no bearing on the model's token-handling limits. A model with a 100K-token context window can still time out because generation speed, network latency, or throttling, while a fast short-context model may respond in milliseconds. Timeout is about availability and latency, whereas context window is about the transformer's fixed capacity for input plus output sequence length.

• ✗

  The geographic region where the AI model is hosted

  Why it's wrong here

  Choosing an Azure region for deployment determines where your model runs, affecting data residency, network latency, and SLA boundaries; it does not change the model's internal context length. The context window is fixed by the model's architecture and checkpoint (e.g., GPT-4o vs. Llama variants), not by the datacenter location. A model hosted in East US and one hosted in Europe can have identical context windows, so geographic region is an infrastructure detail, not a semantic capacity.

Go deeper

Related to this question