More than a month after its promised June release, Google’s Gemini 3.5 Pro remains unavailable to the public. The flagship AI model, teased at Google I/O 2026 by CEO Sundar Pichai as arriving “next month,” has missed its target amidst internal reports of coding performance shortfalls, leaving Windows developers and IT planners in limbo as competitors surge ahead.

What Exactly Is Missing

At Google I/O 2026 in May, the company launched Gemini 3.5 Flash—a faster, lower-cost model for everyday tasks—and promised that a more capable Gemini 3.5 Pro would follow in June. As of July 18, that model has not appeared in the Gemini app, Google AI Studio, or Vertex AI. The public-facing status is unchanged: Gemini 3.1 Pro remains the generally available Pro-class option for cloud and AI users.

On July 17, Bloomberg reported that Gemini 3.5 Pro was months behind schedule because Google was still working to improve its coding capabilities. Reuters corroborated that the model had fallen short of internal benchmarks. In response to both outlets and to Mashable, Google issued a carefully worded statement: “We’re currently testing 3.5 Pro, an upgraded Flash model, and other models with partners, and we’re productively engaged with the U.S. government on model testing and broader frameworks.” The company emphasized cost-effectiveness and shipping speed across its model range, but offered no revised launch date.

The core fact for anyone tracking Google’s AI roadmap is simple: Gemini 3.5 Pro is in testing, not in broad release, and the June window has passed with no public commitment to a new timeline.

What This Means for Windows and Enterprise Users

For most home Windows users, the delay is a non-event. The Gemini consumer experience—whether through the web app, Chrome integrations, or Android—currently relies on Flash and older Pro models, and nothing breaks because 3.5 Pro isn’t here. Unless you’re waiting for a specific advanced feature that was tied to the new model, your workflow remains unchanged.

For power users and developers, the impact is more concrete. Gemini’s models are increasingly woven into tools that Windows-based IT teams use daily: code generation in Android Studio and Cloud Code, document analysis in Workspace, automation via Vertex AI Agent Builder, and API integrations for custom applications. A stronger Pro model with superior reasoning and coding capabilities was expected to unlock more reliable agent workflows, complex refactoring tasks, and enterprise-grade security analysis. Those gains are now on hold.

For IT administrators and enterprise architects, the delay is a planning headache. Many organizations evaluate AI models months ahead of deployment. Teams that penciled in Gemini 3.5 Pro for pilot programs based on the June promise must now either extend timelines or pivot to available alternatives. The key takeaway: do not build deployment plans around conference-stage timelines. The only models you can safely evaluate today are those with published service level agreements and general availability.

For organizations using Google Cloud and Vertex AI, the absence of a frontier Pro model creates a competitive gap. While Gemini 3.5 Flash has gained computer-use capabilities and remains cost-efficient, it is not designed to rival the top-tier reasoning performance of Anthropic’s Claude Mythos Preview, OpenAI’s GPT-5.6 Sol, or the newly open-sourced Moonshot Kimi K3. Enterprises that need cutting-edge code generation or security analysis may need to test competing models in the interim.

A Timeline of Stumbles and Shifting Roadmaps

To understand how we got here, it helps to rewind to the signals and promises leading up to the delay:

  • May 2026: At Google I/O, Sundar Pichai announces Gemini 3.5 Flash and tells a media briefing that 3.5 Pro is “showing great improvements” and will “roll out to everyone next month.” No caveats about testing phases are publicly emphasized.
  • June 2026: Gemini 3.5 Flash expands with computer-use capabilities, enabling the model to take control of a desktop browser for task automation. The Pro model remains absent; no official update is given.
  • July 9, 2026: OpenAI launches GPT-5.6 Sol with advanced cybersecurity coding abilities, raising the bar for coding-focused frontier models.
  • July 17, 2026: Bloomberg and Reuters report that Gemini 3.5 Pro’s coding performance has not met internal goals, causing frustration inside Google. Anthropic’s Claude Mythos Preview and Moonshot’s Kimi K3 have already leapfrogged Google in some benchmarks. Google’s statement acknowledges testing but gives no release date.
  • July 18, 2026: Marshall’s original report notes that the model is still unavailable more than a month after the promised window.

The competitive context is brutal. Since May, Anthropic released its most advanced model yet (Claude Mythos Preview, later a version called Fable 5), OpenAI shipped GPT-5.6 Sol, and Chinese lab Moonshot released Kimi K3, a 2.8-trillion-parameter open-source model that early testers compare favorably to the paid contenders. Google, with its vast resources and distribution power, has found itself playing catch-up on the leaderboard.

What to Do Now: Actionable Steps

If you’re building or planning on Windows with Google’s AI tools, here’s how to navigate the uncertainty:

  1. Evaluate Gemini 3.5 Flash as your default. It’s generally available, cost-effective, and now includes computer-use capabilities. For many code-generation, summarization, and API-driven tasks, Flash will suffice. Start your proofs of concept there rather than waiting for Pro.

  2. Lock in on Gemini 3.1 Pro for production workloads. If you need a Pro-class model today, 3.1 Pro remains the supported, generally available option. It has documented pricing, SLAs, and stability guarantees that an unreleased model cannot match.

  3. Do not plan infrastructure around an unshipped model. Any deployment timeline that assumes Gemini 3.5 Pro will be available by a specific date is speculative. Factor the model into your roadmap only after Google announces general availability and you’ve completed internal validation.

  4. Test competing models if coding or security reasoning is critical. With GPT-5.6 Sol and Claude Mythos already shipping advanced coding features, it may be prudent to run side-by-side evaluations. Windows-based development environments can integrate with these APIs just as easily as Gemini, so a vendor-agnostic evaluation framework is low-risk.

  5. Watch for an upgraded Flash model. Google’s statement explicitly mentions testing “an upgraded Flash model.” A more capable Flash variant could arrive before Pro and serve as a stepping stone.

  6. Stay tuned to Google’s public channels. Official announcements will first appear on the Google AI blog and the Google Cloud blog. Set up alerts rather than relying on social media rumors.

Outlook: What’s Next for Google’s AI Lineup

Gemini 3.5 Pro is not canceled. Google’s statement, while non-committal, confirms active testing with partners. The delay suggests that the model’s coding capabilities are being overhauled—possibly to match or exceed what rivals are now offering. When it finally launches, it may be stronger than the version that would have shipped in June.

In the meantime, the AI landscape will not stand still. Every week a new model or feature resets expectations. For Windows developers and IT teams, the practical lesson is timeless: build on what’s available, evaluate on real tasks, and treat conference promises as aspirations until the bits land in your hands.