A developer is building a chat interface on top of a Foundry-hosted GPT model. User testing shows that although the model typically finishes a full response in about six seconds, people perceive the assistant as slow and sometimes leave before the answer appears, because the message stays blank until the entire response is ready. Which change would most directly address this perception without changing the model or the total generation time?
Select an answer to reveal the explanation.
Short Explanation
The real problem here is not that the model is slow, it is that people are staring at nothing for six seconds and assuming it is broken. The fix is to change how the answer gets delivered, not how fast it gets generated: instead of holding the whole response until it is completely finished and dumping it on the screen all at once, send each piece of text to the screen the moment it is produced, so words start appearing almost instantly and keep flowing in. That single change makes the same six-second response feel responsive instead of frozen. Shortening the answers just trades one problem for a smaller version of the same problem, since people would still wait for a block of text to show up all at once. Caching only rescues the rare case where someone asks the exact same question as someone before them, which will not cover most real conversations. And upgrading to reserved capacity is about handling more traffic reliably, not about making the first words show up sooner for any individual request.
Full Explanation
The correct answer is D. Streaming changes how the response is delivered, sending tokens to the client as soon as the model generates them, so the user sees the answer building word by word within a fraction of a second rather than staring at a blank screen for six seconds; this directly fixes the perceived-latency problem without shortening the response or altering how fast the model itself works. Option A is incorrect because reducing max_tokens shortens the content of the answer, potentially cutting off useful information, and it still leaves the user waiting for a full block of text to appear at once, just a smaller one. Option B is incorrect because caching only helps when a question has been asked before and the exact response can be reused; it does nothing for the large share of unique customer questions that make up normal chat traffic. Option C is incorrect because provisioned throughput is about reserving consistent capacity to handle a predictable request volume reliably, an infrastructure and cost decision, not a mechanism for displaying partial output sooner; it would not change when the first visible text appears to the user.