谷歌Gemini 1.5 Pro人工智慧模式發布，為前所未有的資訊處理提供動力

推出 Google 突破性的 Gemini 1.5 Pro AI 模型，這是 AI 能力的革命性飛躍。憑藉 100 萬個令牌上下文窗口，Gemini 1.5 Pro 可以根據前所未有的資訊量處理和產生回應。該人工智慧建立在專家混合架構之上，既高效又多功能，與前輩相比，表現出顯著的性能改進，勝率高達 87.1%。

Google Unveils Groundbreaking "Gemini 1.5 Pro" AI Model, Empowering Unprecedented Information Processing Capabilities

谷歌推出突破性的「Gemini 1.5 Pro」人工智慧模型，賦予前所未有的資訊處理能力

Mountain View, California - Google has unveiled a transformative advancement in artificial intelligence technology with the introduction of its Gemini 1.5 Pro model. This state-of-the-art AI system represents a quantum leap forward, offering unparalleled capabilities for processing and understanding immense volumes of information.

加州山景城 - 谷歌推出 Gemini 1.5 Pro 型號，展示了人工智慧技術的革命性進步。這種最先進的人工智慧系統代表了一次巨大的飛躍，為處理和理解大量資訊提供了無與倫比的能力。

Central to the Gemini 1.5 Pro's extraordinary capabilities is its astonishing context window, spanning an unprecedented 1 million tokens. This signifies the AI's ability to simultaneously consider a vast amount of data, far surpassing the capacity of its predecessors. This enhanced context window empowers the model to process, comprehend, and generate responses based on comprehensive datasets, encompassing lengthy written documents, intricate codebases, and even multimedia content such as videos and audio files.

Gemini 1.5 Pro 非凡功能的核心是其驚人的上下文窗口，涵蓋了前所未有的 100 萬個代幣。這意味著人工智慧有能力同時考慮大量數據，遠遠超過其前輩的能力。這種增強的上下文視窗使模型能夠基於全面的資料集處理、理解和產生回應，這些資料集包括冗長的書面文件、複雜的程式碼庫，甚至是視訊和音訊檔案等多媒體內容。

Underpinning the Gemini 1.5 Pro model is a novel architecture known as Mixture-of-Experts (MoE). This advanced architecture enhances the model's efficiency during both training and deployment. The MoE architecture enables the AI to selectively process information, activating only the most relevant portions of its neural network based on the input it receives. Consequently, the model operates with greater speed and energy efficiency.

Gemini 1.5 Pro 模型的基礎是一種稱為 Mixture-of-Experts (MoE) 的新穎架構。這種先進的架構提高了模型在訓練和部署過程中的效率。 MoE 架構使人工智慧能夠選擇性地處理訊息，根據接收到的輸入僅激活神經網路中最相關的部分。因此，此模型的運行速度更快，能效更高。

The quantum leap in performance achieved by the Gemini 1.5 Pro model is evident in its benchmark results. Compared to the previous Gemini 1.0 Pro model, the 1.5 Pro boasts an impressive 87.1% win rate. Furthermore, it outperforms the 1.0 Ultra model by a remarkable 54.8% across a spectrum of benchmarks encompassing mathematics, science, reasoning, multilingualism, video understanding, and code. These results underscore the model's versatility and competence in handling complex information processing tasks.

Gemini 1.5 Pro 型號所實現的效能飛躍在其基準測試結果中顯而易見。與之前的 Gemini 1.0 Pro 型號相比，1.5 Pro 擁有令人印象深刻的 87.1% 勝率。此外，在數學、科學、推理、多語言、視訊理解和程式碼等一系列基準測試中，它的性能比 1.0 Ultra 模型高出 54.8%。這些結果強調了該模型在處理複雜資訊處理任務方面的多功能性和能力。

Beyond its raw computational capabilities, the Gemini 1.5 Pro model's accessibility is a key factor in its appeal for developers and enterprises alike. Google provides a limited preview of this AI powerhouse through its AI Studio and Vertex AI services. This preview is offered free of charge, although users may experience increased latency due to the experimental nature of the feature. Google intends to introduce tiered pricing models based on context window size in the future, ensuring scalability and adaptability to meet diverse requirements and budgetary constraints.

除了原始運算能力之外，Gemini 1.5 Pro 模型的可近性也是其吸引開發人員和企業的關鍵因素。谷歌透過其 AI Studio 和 Vertex AI 服務提供了這個 AI 巨頭的有限預覽。此預覽版是免費提供的，但由於該功能的實驗性質，用戶可能會遇到延遲增加的情況。谷歌打算在未來推出基於上下文視窗大小的分層定價模型，確保可擴展性和適應性，以滿足不同的需求和預算限制。

The release of the Gemini 1.5 Pro model positions Google as an industry leader in AI innovation, challenging competitors and establishing a new benchmark for artificial intelligence capabilities. This cutting-edge technology not only exemplifies Google's unwavering commitment to pushing the boundaries of AI but also reflects its dedication to democratizing access to advanced AI solutions.

Gemini 1.5 Pro型號的發布使Google成為人工智慧創新領域的行業領導者，挑戰競爭對手，樹立人工智慧能力的新標竿。這項尖端技術不僅體現了Google對突破人工智慧界限的堅定承諾，也體現了其致力於實現先進人工智慧解決方案民主化的承諾。