After the release of Gemini 3, Gemini 3 Pro achieved a score of 1501 Elo on the LMArena public leaderboard, setting a new historical high, surpassing GPT-5.1, Claude 4.5, and Grok-4.1, becoming the highest-rated multimodal model to date.
In terms of performance, Gemini 3 Pro scored 37.5% on the "Human Ultimate Exam," 91.9% on GPQA Diamond, and 81% and 87.6% on MMMU-Pro and Video-MMMU respectively, demonstrating its comprehensive superiority in scientific, mathematical, and video understanding tasks. The enhanced reasoning mode Deep Think further increased the score on the "Human Ultimate Exam" to 41%, and achieved 45.1% on ARC-AGI-2, setting a new record for general intelligence evaluation.
After the model's release, Sam Altman, CEO of OpenAI, posted on the X platform, "Gemini 3 looks very good," and Sundar Pichai, CEO of Google, responded with "🙏"; Elon Musk also left a message on the official LMArena account, "It's truly worth celebrating," and revealed that Grok 4.20 will be released soon.
In the market, Altman admitted in a leaked internal memo that Google's rapid progress may bring "temporary economic challenges" to OpenAI, and reminded employees that future external evaluations would be "quite severe," indicating that the strong performance of Gemini 3 has intensified the AI competition landscape in Silicon Valley.
