Google promised much. At its I/O conference in May 2026, CEO Sundar Pichai told developers that Gemini 3.5 Pro would arrive the following month. The audience groaned when it didn’t debut on stage. Months later, the model remains stuck in testing. Rivals have noticed. And they aren’t shy about it.
Rivals seize on the stumble
“Gemini who?” Business Insider captured Alexandr Wang, Meta’s chief AI officer, posting on X after a leaderboard placed one of his company’s models above a recent Google release. The jab landed. Thibault Sottiaux, a technical staff member at OpenAI, replied to news of Google’s pre-training work on Gemini 4 with a simple line: “Hope it finishes one day too!” The exchanges reflect a shift. Once viewed as a leader, Google now draws mockery for its pace.
But the delay runs deeper than public taunts. Bloomberg reported on July 16 that the company sits months behind schedule. Engineers have spent extra time chasing improvements, especially in coding tasks. Late June brought fresh training data aimed at fixing those weaknesses. Results disappointed. Ten current and former employees described frustration inside the ranks. Many worry the company risks ceding ground to Anthropic and OpenAI.
The pattern shows. Gemini 3.5 Pro missed its June target. It slipped again past a rumored mid-July window. Some accounts now point to late July or beyond. A 9to5Google article from July 16 tied the holdup directly to coding shortfalls. Retraining didn’t close the gap. And the organization itself adds friction. DeepMind, Google Cloud, Android and Search teams build overlapping tools. Priorities clash. Compute resources stay tight. So decisions drag.
Google hasn’t stayed quiet on everything. On July 21 it released updated lightweight models. Gemini 3.6 Flash and Gemini 3.5 Flash-Lite became generally available, according to the company’s own release notes. These bring better token efficiency, stronger code and agentic planning, and lower prices. A Reuters story the same day noted the contrast. The flagship stays delayed while cheaper, faster options ship. Sundar Pichai highlighted the momentum in his Q2 earnings remarks. The Gemini app reached 950 million monthly active users. Model APIs process 22 billion tokens per minute, driven by those Flash variants. Cloud revenue jumped 82 percent. Revenue overall grew 24 percent.
Yet the frontier model matters. Enterprises want it for complex work. Developers benchmark against it. And competitors keep moving. OpenAI shipped GPT-5.6. Anthropic advanced its offerings. Chinese labs dropped powerful open-weight models. Google’s absence from the top tier leaves a vacuum. One analyst told Business Insider the delay “shifted perception from leading edge to trailing edge.” Josh Beck at Raymond James made the call. He sees the rapid industry tempo as the root cause. Others agree the gap isn’t fatal. Anshel Sag of Moor Insights & Strategy noted Google’s size forces a different rhythm. “It’s just way too early to count anyone out,” he said.
Some users already moved on. Kyle Walker, founder of Clearfork Intelligence, praised Gemini 3.5 Flash as his daily driver for agentic document extraction. High-value work doesn’t always demand the biggest model. That fact buoys Google’s strategy. It pushes cost-effective options hard. Token prices matter when usage scales. Enterprises adopt what works now. Ninety percent of the Fortune 100 already use Gemini Enterprise, Pichai said.
Still, questions linger. Talent has left. Reports tie four senior DeepMind researchers to exits around the same period, some heading to Anthropic. Market reaction followed the Bloomberg story. Alphabet shares dropped sharply, erasing roughly $200 billion in value in one session, per multiple accounts. The stock recovered some ground. But the signal registered. Investors watch frontier progress closely.
Google says it tests 3.5 Pro with partners. A spokesperson told reporters the company ships across a wide range of models while keeping them cost effective. No firm launch date has emerged. Pre-training on Gemini 4 has started, a positive sign for longer-term plans. But the immediate absence of 3.5 Pro invites scrutiny on Wednesday’s earnings call. Pichai will likely face direct questions.
Coding remains the sore spot. It has become a prime enterprise use case. Models that handle long-horizon reasoning, avoid hallucinations, and deliver consistent outputs win contracts. Google’s late-June data update targeted exactly those areas. When benchmarks fell short again, teams reset. The decision shows caution. Rushing a subpar model would damage trust more than delay. Yet each week without it hands rivals time to pull further ahead.
The broader picture reveals trade-offs. Google built scale across Search, Android, YouTube and Cloud. Those products integrate AI deeply. They generate revenue today. Flash models power much of that activity. The frontier model sells the narrative of leadership. When it lags, perception suffers. Rivals pounce. “Gemini who?” stings because it plays on doubt.
Analysts don’t write Google off. Its resources dwarf most startups. Partnerships with governments and enterprises provide testing ground and feedback. Safety talks with U.S. officials continue, adding another layer of review. The company insists it will release 3.5 Pro when ready. That stance buys time. But the clock ticks louder with every competitor announcement.
So Google doubles down on what it can ship. Faster. Cheaper. Widely available. It bets many customers prioritize practicality over raw power. For now, that approach pays. Cloud growth impresses. User numbers climb. Yet the missing Pro model looms. If it arrives strong, the narrative flips. If further slips occur, the jabs will sharpen.
One thing stays clear. The AI race waits for no one. Google learned that lesson the hard way this summer. Its response in the coming months will decide whether the delay becomes a footnote or a turning point.
WebProNews is an iEntry Publication