Gemini 3 Pro Preview from Google DeepMind is now the leading AI model, surpassing OpenAI's GPT-5.1 by 3 points in the Artificial Analysis Intelligence Index. This model excels in five out of ten evaluations, including significant benchmarks like GPQA Diamond and SciCode. Notably, it scored 37% on Humanity's Last Exam, an improvement of over 10 percentage points from the previous best. Its performance in the AA-Omniscience evaluation highlights its capability to provide accurate answers while minimizing errors, suggesting that Gemini 3 Pro is likely a larger model than its competitors. In terms of coding efficiency, Gemini 3 Pro shows impressive results, leading two out of three coding evaluations with a score of 56% in SciCode. It also demonstrates strong performance in agentic contexts, ranking high in Terminal-Bench Hard and Tau2-Bench Telecom. The model's multimodal capabilities allow it to process text, images, video, and audio, making it versatile in various applications. On the MMMU-Pro benchmark for reasoning with images, it outperformed all other models. Despite these advancements, Gemini 3 Pro comes with a high price tag, costing $2 to $12 per million input/output tokens for contexts under 200,000 tokens. This pricing structure drives a 12% increase in operational costs compared to its predecessor, Gemini 2.5 Pro. For larger contexts (200,000 tokens or more), the costs rise further to $4 to $18 per million tokens. Speed remains competitive, with an output rate of 128 tokens per second, aided by Google's TPU accelerators. Overall, Gemini 3 Pro combines advanced intelligence with high operational costs, making it a significant player in the AI field.