Gemini 3.5 Pro still not out as its 'next month' window closes

At Google I/O on 19 May 2026, Google launched Gemini 3.5 Flash — fast, cheap, and immediately the default in the Gemini app and Search's AI Mode — and announced Gemini 3.5 Pro without releasing it, saying Pro was being used internally and would roll out 'next month'. Said in mid-May, that pointed squarely at June. By late June, the window had effectively closed: Pro was not in the consumer Gemini app, not in Google AI Studio (where developers test models), and not in the stable API. Flash remained the 3.5 model people can actually use today, and the only one with a published model card.

Two cautions. First, the impressive specs in circulation — a ~2-million-token context and a 'Deep Think' mode — are trade-press expectations, not Google-confirmed figures for 3.5 Pro; Flash's confirmed context is about 1M tokens. Second, a late flagship is normal: larger models usually trail their faster siblings because evaluation and serving take longer, so the slip mainly shows the 'next month' framing was optimistic, not that the model is troubled. The signal that Pro has gone from announced to usable is concrete — a stable API model ID and a published model card with benchmarks. Neither existed as of late June.