Google’s next flagship model—gemini-3.5-pro—briefly appeared on Arena.ai (LMSYS Chatbot Arena) for live A/B testing, sending ripples through the tech and AI communities.
Though the model was pulled in less than an hour, the sudden test sparked intense speculation across X and Discord. As covered in recent coverage of Google’s AI model updates on NokiaPowerUser, this unexpected staging signals that the search giant is preparing its next major AI push.
Here is a breakdown of what happened, how early testers reacted, and what this signals for Google’s upcoming release schedule.
What Happened on Arena.ai?
Community members conducting blind prompt battles noticed gemini-3.5-pro popping up in active testing pools late yesterday.
Key details from the brief leak include:
Direct Confirmation: When prompted, the model identified itself outright: “I am Gemini, a large language model built by Google.”
Concurrent Glitches: The appearance coincided with brief micro-outages and API hiccups across Google’s developer infrastructure, pointing to backend staging.
A Quick Exit: Within 60 minutes, automated tracking bots confirmed the model had been removed from rotation.
New Arena Models
• gemini-3.5-pro
How Are Early Testers Reacting?
Because LMSYS relies on blind preference scoring, no official benchmark rankings or Elo scores are available yet. As noted in ongoing reports on Gemini model developments at NokiaPowerUser, early tester sentiment remains split between anticipation and skepticism:
The Positives
High Expectations for Coding & Multimodality: Users anticipate major generational leaps in agentic coding, visual UI analysis, and complex multi-modal capabilities.
Pre-Launch Indicator: Historically, unannounced appearances on Chatbot Arena serve as final pre-flight stress testing, often indicating a public rollout is imminent.
The Concerns
Underwhelming Initial Outputs: Some early testers felt the responses were incremental rather than revolutionary, raising questions after months of anticipation.
Uncalibrated Builds: Experts emphasize that early, uncalibrated test models on Arena frequently perform differently than polished production APIs.
OverviewFeatureDetailsModelgemini-3.5-proPlatformArena.ai (LMSYS Chatbot Arena)Official BenchmarksPendingTest WindowUnder 1 hourExpected UpgradesReasoning, agentic coding, native multimodalityCould Gemini 3.5 Pro Launch Next Week?
Google has held onto its next-generation Pro model longer than usual. For the latest developments and real-time updates on Google’s releases, check out NokiaPowerUser’s Android and Google news hub.
While there is no official confirmation, rapid deployment and withdrawal on Arena.ai traditionally indicate final system checks. Developers and tech enthusiasts should keep a close eye on Google DeepMind announcements over the coming days.
Add NPowerUser as a preferred source on Google News