Google was supposed to launch Gemini 3.5 Pro in June. Then July. Latest reports point to July 17. Sundar Pichai said at I/O in May: "Give us until next month to get it to you." The audience groaned. A month passed. Nothing.
Google has done this before. Rushed models. Underwhelmed developers. Fell behind OpenAI and Anthropic on the one thing that matters most to enterprises: coding.
Pichai himself admitted it. In a rare moment of public honesty, he acknowledged that Google is lagging in AI programming for complex software tasks.
Gemini 3 Pro had a nasty habit of "slacking off" — automatically rewriting code too aggressively, refusing to dive deep when the task got complicated. Developers hated it. Internal staff posted anti-AI memes mocking Google's own coding tool.
The delay isn't about one bug. It's about a pattern.
Early testers flagged two big problems with Gemini 3.5 Pro. First, token inefficiency: the model burns through tokens too quickly, especially on programming tasks. Second, token inefficiency means higher cost — and in enterprise AI, cost is the vote that matters most.
Google also reportedly abandoned its original Gemini 2.5 Pro base and started fresh, with a new pre-training run. That's expensive. That's time-consuming. That's also the kind of decision you make when you want to ship something that actually competes.
The leaked specs are still impressive. Two-million-token context window. "Deep Think" reasoning mode. SWE-Bench Verified score of 80.6%, close to GPT-5.5's 82.7%. Multimodal coding that understands screenshots and error logs directly.
But the market has changed. OpenAI just dropped GPT-5.6. Anthropic has Fable 5. Both are already in production. Both are winning developers.
Google is late. Again.
But maybe this is the moment Google stops being late and starts being deliberate. The delay gives them time to fix token efficiency. It gives them time to fix the "laziness" issue. It gives them time to price correctly — reports suggest a premium-tier positioning, above Gemini 3.1 Pro but justified by performance.
Google's stock has underperformed the Nasdaq AI index by 11% this year. The pressure is real.
If Gemini 3.5 Pro ships on July 17 and delivers, the delay becomes a footnote. If it ships and still lags behind, the delay becomes the story.
P.S. If you're a developer waiting for Gemini 3.5 Pro, Google is doing the right thing by not shipping broken AI. But Fable 5 and GPT-5.6 aren't waiting — and neither are your cloud budgets.
