GCP Generative AI Leader Practice Question 63: Flash vs Pro Cost Latency
Google Cloud Generative AI Leader · Test 7 · question 3 of 10
A mobile app team is adding an in-app feature that classifies incoming support messages into a handful of categories and must respond in under a second, at high volume, and at the lowest possible per-call cost. Response quality only needs to be good enough to route the message correctly, not to write long, nuanced prose.
Which Gemini model family best fits this requirement?
Show answer and explanation
Answer:
Google positions the Flash variant of Gemini as the lightweight, low-latency, low-cost option in the family, purpose-built for high-throughput tasks like classification and routing where response time and price per call matter more than deep reasoning or long-form generation. That matches the sub-second, high-volume, low-nuance requirement in the scenario exactly.
- A is wrong because a large context window solves a different problem, handling long inputs, not the stated need for speed and low cost on a simple classification task. - B is wrong because reaching for the most capable, most expensive model for a task that just needs "good enough" routing is over-provisioning that increases both cost and latency. - D is wrong because it overgeneralizes: self-hosting an open model has its own infrastructure and operational cost, and whether it beats a hosted model on cost depends on the deployment and traffic pattern, not a blanket rule.
Reference: Gemini 2.5 on Vertex AI: Pro, Flash and Model Optimizer live