The release of the next generation of Google Frontline AI Model Gemini 3.5 Pro was allegedly postponed from June to July. Business Insider cites sources who claim that Google is continuing to adapt the model based on feedback from early testers, thus delaying the go-live.
At the I/O Conference of Developers in May this year, Google has forecast this new model, but it was not officially opened. Google CEO Pichaipei said that the model would be launched “in the next month”. The latest release time changes were not commented on by Google speakers.
Reasons for delay and reorientation
It was reported that the delay was not a change of route, but rather to allow more time to collect data on the actual use of the scene. According to the source, Gemini 3.5 Pro has been made available to some users via the Google Antigravity platform and also appears on the AI benchmarking website LMArena.
The focus directions of the new model include greater long-cycle task-processing capacity and support for AI proxy applications. Such capabilities are becoming a focus of competition for large models as enterprises become more concerned with the scenes of automatic implementation and multistep collaboration.
Increased competition for coding capabilities
The report mentions that, although Gemini 3 performed well beyond some market expectations last year, OpenAI and Anthropic continue to widen the gap between Google in the clearest AI use of coding. Gemini 3.5 Pro's roll-out rhythm was also observed in a more competitive context.
Flash 3.5 feedback incorporated into improvements
According to the source, Google has integrated feedback from the recent Flash 3.5 model into the 3.5 Pro optimization process. One of the criticisms is that Flash consumed too fast in its use.
This means that Gemini 3.5 Pro ' s adjustment is not just about capacity enhancement, but also about cost and efficiency performance. For business clients, the stability of models in complex missions and the efficiency of token use affect actual willingness to deploy.