Gemini 3.5 Pro Misses Its Third Launch Deadline as Google Eyes a Stopgap Release

My take: Gemini 3.5 Pro did not launch today. The model has now missed three target dates: May, June, and now July 17 — the same day the WAIC conference opened in Shanghai with hundreds of new AI products on display. According to reports, the issue is hallucinations and reliability gaps Google has not yet been able to resolve, and the company is evaluating a limited stopgap release while the full model continues in development.

This is not a story about a Google-specific failure: it is what it costs to build at the AI frontier in 2026, when GPT-5.6, Grok 4.5, and Kimi K3 are already in production. The margin for reliability errors at this level is zero.

For any professional choosing AI tools, the practical lesson is this: lab launch dates are commitments that move. It is better to evaluate a model when it is actually available and test it against your real use cases, than to wait for the next launch to change everything before you start working with AI.

Do you have a systematic way to evaluate new models when they release, instead of assuming the most recent one is always the best fit for your work?