Gemini 1 arrived with a multimodal reset

Gemini 1.0 changed the shape of the model race by making multimodality and model placement part of the launch.

Ultra, Pro, and Nano implied different jobs. Intelligence could live on the device, in the normal product route, or behind an expensive frontier call. The architecture no longer had to pretend one model was the right answer for every task:

small on-device model -> quick local work
pro model -> normal multimodal product flow
ultra model -> hard reasoning or expensive review

Mixed input also stopped feeling like an exception. A screenshot, voice note, chart, and document could belong to the same request. That is useful only if the product keeps those inputs together long enough for the model to see their relationship.

A billing flow might look like:

user uploads a screenshot of a charge
user says: "this is not mine"
client keeps image, visible text, and account context together
pro route handles the first answer
ultra route handles unresolved fraud or duplicate-charge cases

The model tier does not remove the routing work. The product still has to decide which requests stay local, which belong in the normal app path, and which justify the expensive route.

Gemini 1.0 made that split visible early. Once intelligence has a location, model selection becomes part of product architecture.