Gemini 3.5, the new model family Google unveiled at Google I/O in May, began rolling out with the faster 3.5 Flash. But the top-tier 3.5 Pro has still not reached general availability, even though Google originally pointed to "the following month." Google's official page still shows a "coming soon" label. Here is what we know so far, based on official information.

Where the Gemini 3.5 Series Stands Now

Google introduced Gemini 3.5, its latest model family, at Google I/O on May 19, 2026. The series launched first with the faster 3.5 Flash, which became available across a wide range of surfaces: the Gemini app, AI Mode in Google Search, the developer-facing Google AI Studio and the agent development platform Google Antigravity, and Gemini Enterprise for businesses.

For the top-tier 3.5 Pro, Google said at the time that it was "already being used internally" and that it planned to "roll it out next month." Yet as of mid-July 2026, nearly two months after the announcement, Google's official model page still marks 3.5 Pro as "coming soon." The model generally available in the Pro tier remains the previous-generation Gemini 3.1 Pro (Preview). Gemini 3.1 Pro accepts up to 1 million input tokens and up to 64,000 output tokens, and supports image, audio, video, and PDF input in addition to text.

What the Earlier "3.5 Flash" Signals for the Pro Version

We have not yet seen what 3.5 Pro looks like, but the earlier 3.5 Flash offers a sense of what the series prioritizes. Google positions 3.5 Flash as a model strong in long-horizon agentic work (processing that plans and carries out multiple steps on its own) and in coding.

By Google's own figures, 3.5 Flash surpasses the previous-generation 3.1 Pro on practical, hands-on benchmarks, scoring 76.2% on the agent-oriented Terminal-Bench 2.1 and 83.6% on MCP Atlas, which measures tool use. Its output token throughput per second is said to reach roughly 4 times that of other frontier models. In practice, companies such as Shopify, the Australian financial group Macquarie Bank, Salesforce, Ramp, Xero, and Databricks have begun building 3.5 Flash into their work for automation and large-scale data analysis. The new AI agent Gemini Spark, which handles personal schedules and tasks on the user's behalf, also runs on 3.5 Flash.

That said, these figures and case studies were all published by Google itself. Real-world performance varies with the task and the conditions, so it is best to weigh them alongside third-party verification. Once 3.5 Pro arrives, it is expected to build on this foundation to handle even harder reasoning and longer multi-step work.

Reports on Timing vs. Google's Cautious Stance

As for when 3.5 Pro will reach general availability, some overseas media reports have described a target of mid-July. However, Google has not officially confirmed that date, and it has yet to publish a model card, pricing, or official benchmarks for 3.5 Pro. The only thing that can be stated with certainty right now is that a rollout originally described as coming "the following month" has slipped by several months.

Part of the backdrop is that competition among frontier models has intensified further. As OpenAI and Anthropic push out new models one after another, Google appears to be filling the gap in the Pro tier with the current 3.1 Pro while it finishes 3.5 Pro. For developers and users, the prudent approach is to plan around 3.1 Pro until a firm date appears, and to wait for the official announcement.

Summary

In the Gemini 3.5 series, 3.5 Flash shipped first, while the top-tier 3.5 Pro remains parked at "coming soon" even after the original "following month" guidance has passed. There are reports of a mid-July rollout, but Google has not yet provided an official date or model card. How far 3.5 Pro will push the agentic and coding-focused direction that 3.5 Flash has shown remains to be seen. We will be watching for the official announcement.