Google’s long-awaited **Gemini 3.5 Pro** AI model appears to have made a brief and unexpected appearance on the **Arena** benchmarking platform before being removed just minutes later, fueling speculation that the company is preparing for a wider public release after months of delays.
Several users on **X (formerly Twitter)** reported spotting the new model on the AI evaluation platform on **July 31**, sharing screenshots that suggest Google briefly enabled access before taking it offline.
## Brief Appearance Lasted About 30 Minutes
Among the first to report the incident was X user **Harshith (@HarshithLucky3)**, who claimed that **Gemini 3.5 Pro** remained available for only around **30 minutes** before disappearing from the platform.
According to the shared screenshots, the user was able to complete only a single SVG-related benchmark before the model was withdrawn. Other users also posted similar screenshots and confirmed that they had briefly seen the model listed on Arena before it was removed.
While Google has not commented on the incident, the brief appearance has sparked speculation that the listing may have been part of an internal test or an accidental early rollout.
## What Is Arena?
Formerly known as **LMArena**, **Arena** is one of the most widely recognized community-driven AI benchmarking platforms.
Unlike traditional benchmark suites, Arena relies on **blind side-by-side comparisons** conducted by real users. Participants evaluate responses from anonymous AI models across a variety of tasks, including:
* General conversation
* Creative writing
* Programming and code generation
* Mathematics
* Visual and SVG generation
* Reasoning and problem-solving
The platform ranks models using an **Elo-style rating system**, similar to those used in competitive chess, allowing users to compare overall model performance based on real-world interactions rather than synthetic benchmark scores.
Because of its crowdsourced evaluation approach, Arena has become an influential reference point for measuring the capabilities of leading AI models from companies such as Google, OpenAI, Anthropic, xAI, and Meta.
## Gemini 3.5 Pro Has Faced Multiple Delays
Google officially introduced the **Gemini 3.5** family during its developer announcements on **May 19, 2026**.
At launch, the company released **Gemini 3.5 Flash**, positioning it as a fast and efficient model optimized for low-latency applications. At the same time, Google confirmed that **Gemini 3.5 Pro** was already being used internally and was expected to become publicly available the following month.
However, the rollout did not proceed as planned.
Over the following months, the expected launch was repeatedly postponed. Google offered few details beyond stating that the model was undergoing additional testing with selected partners before a broader release.
The brief appearance on Arena is the first public indication that Gemini 3.5 Pro may be nearing wider availability.
## Anticipation Continues
Although Google has yet to announce an official release date, the temporary listing suggests that development and validation of Gemini 3.5 Pro are progressing.
If launched as expected, Gemini 3.5 Pro is likely to become Google’s flagship general-purpose AI model, offering stronger reasoning, coding, and multimodal capabilities than the existing Gemini 3.5 Flash variant.
For now, users will have to wait for an official announcement, but the unexpected appearance on Arena has renewed expectations that Google’s next-generation AI model could arrive sooner rather than later.