Exercise: match five requests to a model family

Original fictional scenario. No cloud account, API calls or paid services are required. Difficulty: intermediate · Estimated duration: 15–20 minutes

A consumer-electronics company receives five requests in one week. All names and details below are invented.

#	Request	Hard requirement
1	Field technicians need an assistant on rugged tablets in basements with no signal	Must run on the device, offline
2	Marketing wants a new logo with the brand's exact lettering	Legible typography
3	Social media wants a vertical product video for phones	30 seconds long
4	Support wants a system that reads warranty photos and writes a claim summary	Image in, text out
5	Design wants to refine a poster over several rounds of spoken feedback	Conversational editing

Your decision

Choose a model family for each request: Gemini, Gemma, Imagen or Veo.
For each choice, name the one fact you would confirm on the model's own page before committing.
Request 3 cannot be met by one clip as stated. Explain why and propose a plan.
Request 2 points to Imagen's strengths. State what Google's current image documentation says that affects a new project, and decide.

Rubric (10 house points)

4 points: a defensible family for each of requests 1, 3, 4 and 5.
2 points: a specific page fact to confirm for each.
2 points: request 3 answered from Veo 3.1's documented clip lengths.
2 points: request 2 decided with both Imagen's strengths and Google's current recommendation.

Reference solution

1 — Gemma. It runs on your hardware and mobile devices; confirm a size small enough for the tablet, since lower parameter sizes need fewer resources. 4 — Gemini. It is built for multimodality across images and text; confirm the chosen model's page lists image input. 5 — a Gemini image model, which is uniquely capable of multi-turn conversational editing. 3 — Veo, which generates video from text prompts and images.

Request 3: Veo 3.1's page lists clip lengths of 4, 6 or 8 seconds and 9:16 for vertical video, so a single 30-second clip is not available as stated. Plan four 8-second vertical clips edited together, or shorten the brief. Rejected alternative: "Veo 3.1 is the latest, so it will do 30 seconds".

Request 2: typography and logos are documented Imagen strengths. But Google recommends Gemini as the starting point for generating images and lists Imagen endpoints such as imagen-4.0-generate-001 as deprecated, with gemini-2.5-flash-image as the replacement. Start the new project on a Gemini image model and test the lettering; treat Imagen as a documented option, not a new dependency.

Sources and scope

All names, requests and rubric points above are house-authored. The model capabilities applied are grounded in:

Match five requests to a model family — decision exercise

Exercise: match five requests to a model family

Your decision

Rubric (10 house points)

Reference solution

Sources and scope