Gemini 3.5 Pro Still in Preview Entering July's Second Week With No Confirmed Launch Date

My take: Gemini 3.5 Pro entered the second week of July the same way it entered the first: in limited preview, without a confirmed general availability date, no benchmarks published by Google, and no final pricing. The model, which promises a 2 million token context window and extended reasoning with Deep Think, is available only to a small group of enterprises in Vertex AI preview and on LMArena for community testing. Google originally had it on the June calendar, then pushed it to July, and July is now underway with no launch.

The practical point for anyone evaluating it: a model in preview is not a model in production. It has no service level agreement, no final pricing, and it can change before general availability. Planning production workflows around a preview version means accepting unnecessary continuity risk.

What this pattern shows, beyond the Gemini case specifically, is that AI lab launch dates are more a directional signal than a commitment. The practical lesson is to evaluate models against what is in general availability today, and to build architectures that allow swapping models without rebuilding all the logic. Do your AI projects depend on a model that does not yet exist in production?