TL;DR<\/p>\n
Gemini 3.5 Pro: Google missed its June rollout window, and the model may be delayed by coding performance below internal expectations. Training Update: Google changed Gemini\u2019s training data in June, but results may not have reached its internal bar. Partner Testing: Google is testing the model with partners but has not provided a confirmed cause or replacement date. Market Pressure: Alphabet shares fell nearly 3% while OpenAI\u2019s GPT-5.6 Sol and GitHub Copilot already serve coding workloads. <\/p>\n
Google has missed the planned June rollout window for its higher-end Gemini 3.5 Pro<\/a> artificial-intelligence model. Its coding performance has reportedly fallen short<\/a> of internal expectations, people familiar with the matter said, though Google has not provided a confirmed explanation or replacement date.<\/p>\n After the missed window, Alphabet shares fell nearly 3%<\/a> as news of the delay emerged. Coding is a competitive workload and a major part of Google\u2019s engineering process, while developers can use rival products as Pro remains unavailable. Each extra month also gives shipping alternatives more time to become established in developer workflows.<\/p>\n During partner testing, an unnamed company spokesperson paired Google\u2019s evaluation of Gemini 3.5 Pro and other models<\/a> with its position that it was shipping quickly across models and keeping them cost-effective for customers. Partner testing can expose problems before broad availability, but it provides neither a launch timetable nor confirmation of the possible cause.<\/p>\n Coding Performance and Google\u2019s Release Process <\/p>\n Google had changed Gemini\u2019s training data in June, while coding remained a focus across generation, understanding and software repair on the flagship model\u2019s path to public release. Results after the training-data update<\/a> appear to not have reached Google\u2019s internal bar for the higher-end model despite the broader family moving forward. Reliable coding requires repository context, tests, dependencies and project rules because a plausible change can still break builds, introduce defects or create more work during human review at repository scale.<\/p>\n Unlike the newer 3.5 models, Pro and Flash became available together<\/a> in 2025, putting both model types into public use during the same release cycle instead of separating the faster and higher-end variants. Gemini 3.5 Flash launched first<\/a>, while Pro was already in internal use in May and remains under limited evaluation, creating a split release sequence without a public date for Pro. Google\u2019s April figures put AI at 75% of all new code, up from 50% the previous fall, making coding quality an operational issue for engineers who approve the output at Google\u2019s scale.<\/p>\n Beyond model quality, people familiar with the matter said shifting mandates and duplicated work, resource competition and stakeholder layers slowed launches while Google Cloud, Google DeepMind and Android are developing separate coding tools. Google is also working to unite its internal systems, requiring teams to reconcile overlapping capabilities, divide computing capacity and decide when a shared tool can enter wider partner testing across all three groups. Across Google\u2019s products, the 2025 Gemini Pro and Antigravity launch<\/a> created a foundation that a successor must improve across Search, Maps and YouTube without compromising their different safety, performance and deployment requirements.<\/p>\n Rivals Are Already Shipping Coding Tools <\/p>\n