Google Delays Gemini 3.5 Pro Launch by Months After Retraining Falls Short

The deadline that slipped

At Google I/O in May 2026, Alphabet CEO Sundar Pichai promised that Gemini 3.5 Pro, the company's frontier model, would ship broadly by June 2026. Since then it has stayed in limited preview. It did not arrive. According to a Bloomberg report cited by 9to5Google on July 16, 2026, Google updated the 3.5 Pro training data in late June in an attempt to fix its coding weaknesses, but internal testing disappointed again. A Google spokesperson confirmed to reporters that the company is currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and is shipping quickly across a wide range of models while keeping them highly cost effective for customers. No new public launch date has been given.

Coding, reasoning, and a data rework that did not close the gap

The reporting behind this story, from Bloomberg, Search Engine Journal, Android Authority, Android Headlines and Technology.org, converges on one diagnosis: the model still falls short on coding tasks and on long horizon reasoning, the ability to hold multiple logical steps coherent across an extended task. The late June training data rework was a direct attempt to fix the first problem, coding, but it solved neither that nor the second. Per Android Authority and Bloomberg, part of the delay has organizational roots: DeepMind, Google Cloud, Android and Search each build overlapping AI coding tools, producing conflicting priorities between teams that slow execution. Compute capacity constraints are cited as an additional factor.

For teams coding and automating with AI, Gemini 3.1 Pro remains the baseline

For SMBs and marketing and customer service automation agencies that rely on frontier models to generate code, integrate systems or orchestrate agents, the practical news is simple: the model in production on Google's Pro tier remains Gemini 3.1 Pro, released in February 2026. There is no indication of when that changes. Any product plan, pricing or commercial offer built on the expectation of a Gemini 3.5 Pro available by June 2026 needs revisiting. The lesson extends beyond Google: frontier model announcements at keynotes and events like I/O are statements of intent, not locked engineering dates. Technical teams that plan product roadmaps on top of launches not yet confirmed in wide preview risk delaying their own deliveries to customers.

Practical implications for choosing a model

In practice, Brazilian agencies and SMBs running customer service and marketing automations on Google's API should keep Gemini 3.1 Pro as the operational baseline until there is a clear signal of general availability for 3.5 Pro, not just partner preview. For AI assisted coding specifically, it is worth testing against alternatives that have already climbed the rankings in the last two weeks, such as OpenAI's GPT 5.6, launched July 9, 2026, and Moonshot AI's Kimi K3, launched July 16, 2026, before assuming Google's next model will fix coding bottlenecks in the short term.

The race got tighter, and Google is playing catch up

The Gemini 3.5 Pro delay is not an isolated stumble, it is a symptom. While Google rewrites training data trying to unlock coding and reasoning, OpenAI and Moonshot AI have already shipped models climbing the competitive rankings. Historically, the edge has favored whoever arrives first with predictable timing, not just whoever arrives with the strongest model. If internal organizational complexity, DeepMind, Cloud, Android and Search competing for priority, keeps outweighing external competitive pressure, Google risks turning its scale advantage into a coordination liability. For the SMB and agency market, the takeaway is twofold: diversify model vendors, and treat every launch date as an estimate, not a commitment.