Gemini 3.5 Pro Slips to July as Google Misses Deadline

Google let its own deadline for Gemini 3.5 Pro come and go. The company's next flagship model was widely expected to reach general availability by the end of June — a target Sundar Pichai all but set himself at Google I/O in May, when he asked the audience to "give us until next month." June 30 arrived with no launch, and Google has now restated that Gemini 3.5 Pro will ship in July. Prediction markets that had been tracking the release moved decisively toward "no June launch" as the month closed.

Google's stated reason is refinement rather than any single blocking bug. According to reporting from TechTimes and others, the company is folding in feedback from testers on its Antigravity platform and the LMArena leaderboard, and is specifically working on token efficiency — a concern carried over from Gemini 3.5 Flash, which some users said burned through tokens too quickly on long prompts and extended tasks. Google is also said to be tuning the model for longer, more agentic workloads, the kind of multi-step autonomy where a small inefficiency compounds into real cost.

Expectations for the model are steep, which raises the bar for getting the launch right. Gemini 3.5 Pro is reported to arrive with a two-million-token context window — double the one-million window on Claude Opus 4.8 and OpenAI's GPT-5.5 — alongside a "Deep Think" reasoning mode and frontier multimodal capabilities. Reported pricing estimates put it around $15 per million input tokens and $60 per million output, which would make it a premium option rather than a budget one, and roughly ten times the cost of Gemini 3.5 Flash. Those figures are unconfirmed, but if they hold, Google would be selling long context and reasoning depth, not a low sticker price.

The slip lands at an awkward moment for Google DeepMind's talent bench. Over the past several weeks the lab has lost a string of senior researchers, including Gemini co-lead Noam Shazeer and Nobel laureate John Jumper — moves BitsMinds covered when two departures landed in three days and when Jumper left for Anthropic. Reports since then describe at least four more senior Gemini researchers heading to Anthropic in late June. None of that necessarily explains the delay, but it deepens the perception problem: Google is asking for more time on its flagship while rivals ship and its people leave.

For now, the frontier's generally available options remain GPT-5.5 and Opus 4.8, with OpenAI's newest GPT-5.6 still in a staggered, government-slowed rollout. Google has shipped aggressively this year — Gemini 3.1 Pro and Ultra, Gemma 4, a raft of Flash variants — but the Pro tier is the one enterprises benchmark against everyone else. Being methodical is defensible. Being late, publicly, is a different kind of cost.