Gemini 3.5 Pro delays due to coding performance, upgraded Flash model in testing (9to5google.com)

Google's highly anticipated Gemini 3.5 Pro, initially set for release in June, is facing delays primarily due to challenges in enhancing its coding performance. Following the mid-May announcement of the Gemini 3.5 Flash model, which was touted for its improvements, the company discovered disappointing results during testing when updating the training data for Gemini 3.5 Pro. Consequently, Google is taking extra time to refine its capabilities, particularly pertaining to coding skills, as they have observed a notable increase in AI-generated code within their organization—from 50% in the fall to 75% by April. This delay is significant for the AI/ML community, as it highlights the complexities involved in developing effective AI coding tools. Google's commitment to upgrade models and expand capabilities amidst these hurdles signals the growing importance of AI in software development. Moreover, the internal push to unify various AI coding tools across different teams at Google, along with the tension between the embrace of AI-generated code and traditional human-written standards, reflects broader trends within the industry as firms seek to balance innovation with rigorous quality control.