A team is building a semantic search feature: users type a question, and the system must find the most similar passages from a document library by comparing meaning rather than exact keywords. They plan to deploy a model from the Foundry catalog to convert both the stored passages and each incoming query into numeric vectors for that comparison. What kind of model should they deploy for this step?
Select an answer to reveal the explanation.
Short Explanation
Semantic search only works if you can turn meaning into something you can measure, like converting sentences into coordinates on a map, where things that mean similar things end up near each other. That conversion, from text into a list of numbers you can compare mathematically, is a distinct job from the job of writing a fluent conversational reply. A model built for chatting is optimized for producing good next words in a dialogue, not for producing a stable numeric fingerprint of a passage's meaning. So even though both are 'language models' in a loose sense, only the one purpose-built to output those numeric fingerprints is the right tool for finding the closest match among millions of passages. Audio and image models are answering an entirely different question, they are not converting text into a comparable representation of meaning at all.
Full Explanation
The correct answer is D. An embedding model is designed specifically to convert text into a numeric vector, and comparing vectors by distance or similarity is exactly how semantic search matches a query to the most relevant passages by meaning rather than exact word overlap. Option A is incorrect because a chat-completion model is built to generate conversational text responses; it does not natively produce a fixed-length vector suited for similarity comparison, and using it for that purpose would be the wrong tool even though it can process language. Option B is incorrect because a text-to-speech model converts text into audio output, which has nothing to do with representing meaning as a vector for comparison. Option C is incorrect because an image-generation model produces pictures from text prompts, a completely different modality with no role in matching query meaning against stored passages.