Gemini 3.5 Pro delay pushes Google launch to July

Google has pushed Gemini 3.5 Pro past the June window it flagged at I/O, with Bloomberg and Business Insider reporting a July target after more work on coding and agent-style tasks. The delay lands in a part of Gemini that Google has been selling to developers and enterprise teams, where coding performance is a buying question rather than a benchmark line.

Google’s public timeline was tighter. In the company’s Gemini 3.5 announcement, it led with the faster 3.5 Flash model, cited a 76.2% Terminal-Bench 2.1 score and said the more capable Pro version would follow the next month. Developers and enterprise buyers were left expecting a short wait between the preview pitch and a broader rollout.

“We’re also hard at work on 3.5 Pro. It’s already being used internally, and we look forward to rolling it out next month.”

Google, in its June Gemini 3.5 announcement

That line now reads softer. Business Insider reported in June that the release had slipped to July as Google kept trying to improve coding capabilities and the way the model handles agentic work. Google has not posted a new launch date in its public Gemini API model documentation, which is still the main reference point for developers checking what is actually available.

Why the delay matters for developers

For Australian software teams weighing Gemini for internal copilots, code generation and workflow automation, the setback is a readiness signal. Buyers can absorb demos and preview claims. Production coding models are judged on whether they can hold context across long edits, call tools reliably and return code that engineers are prepared to ship.

Late model releases are not unusual in AI. A delay tied to coding says more, because that is one of the workloads Google needs to win.

The June announcement also shows how Google is trying to stage the 3.5 family. Flash gives the company a faster model it can point to now, including tool-use claims and benchmark scores. Pro remains the higher-end promise, and that leaves a gap at the top of the range while developers wait to see whether it brings a clear lift in coding performance.

Google has a narrower pitch until Gemini 3.5 Pro is broadly available. Customers are being asked to judge a flagship coding model through company statements, early documentation and reports about internal testing. For developer-focused AI products, that is thin evidence. Enterprise teams, including Australian buyers comparing models for support desks, internal knowledge tools and software delivery, usually want release notes, pricing, testing time and real-world behaviour before they commit.

Google’s own materials still present Gemini 3.5 as a major part of its developer strategy. The July slip puts more weight on coding reliability than model branding. If 3.5 Pro ships soon with stronger performance in the tasks Google has been tuning, the delay may read as a short product-timing problem. Another drift would make the story less about launch choreography and more about how hard frontier AI vendors still find dependable coding work at scale.