Openai揭幕了GPT-4.1，其AI模型陣容中的下一個迭代

這種新模型以GPT-4O的功能為基礎

OpenAI has unveiled GPT-4.1, the latest installment in its renowned large language model (LLM) lineup, during a live broadcast on Monday. The new iteration builds upon the capabilities of GPT-4o, introducing significant upgrades in performance, affordability, and context handling, particularly relevant for developers and enterprise users.

Openai在周一的實時廣播中揭示了GPT-4.1，這是其著名的大型語言模型（LLM）陣容中最新一部分。新的迭代基於GPT-4O的功能，在性能，負擔能力和上下文處理方面進行了重大升級，尤其與開發人員和企業用戶有關。

One of the standout modifications in GPT-4.1 is the expanded context window, now supporting up to 1 million tokens. This adjustment marks a substantial increase from the 128,000-token limit in GPT-4o, enabling the new model to process much larger volumes of text, images, and even videos in a single session. As a result, GPT-4.1 can deliver more accurate and detailed responses for complex tasks involving large datasets or intricate instructions.

GPT-4.1中出色的修改之一是擴展的上下文窗口，現在支持多達100萬個令牌。這種調整標誌著GPT-4O的128,000次限制的大幅增加，使新模型能夠在一次會話中處理大量文本，圖像甚至視頻。結果，GPT-4.1可以為涉及大型數據集或複雜說明的複雜任務提供更準確和詳細的響應。

According to OpenAI, GPT-4.1 also improves reliability, especially in managing both long and short-form content. The model has been fine-tuned to ignore irrelevant data more effectively, which enhances its usefulness across various real-world applications—from coding and instruction-following to data analysis.

根據OpenAI的說法，GPT-4.1還提高了可靠性，尤其是在管理長形和短形式的內容方面。該模型經過微調，可以更有效地忽略無關緊要的數據，從而提高了其在各種現實世界應用程序中的實用性 - 從編碼和指導遵循到數據分析。

In addition to the main GPT-4.1 release, OpenAI introduced two lightweight variants designed for cost-conscious developers:

除了主要的GPT-4.1版本外，OpenAI還推出了針對成本意識開發人員設計的兩個輕量級變體：

* GPT-4.1 Mini – An affordable entry-level model offering core features of the flagship version, ideal for smaller-scale development needs.

* GPT-4.1 MINI - 一種負擔得起的入門級模型，提供旗艦版本的核心功能，非常適合較小規模的開發需求。

* GPT-4.1 Nano – The smallest and fastest variant, built for high-efficiency tasks where speed and cost savings are key.

* GPT-4.1 NANO - 最小，最快的變體，是專為速度和成本節省至關重要的高效任務而設計的。

Notably, GPT-4.1 is priced 26% lower than its predecessor GPT-4o, making it more accessible for startups, enterprises, and independent developers. This pricing shift comes as OpenAI faces mounting competition from rival AI models like DeepSeek, which has recently set a new benchmark for efficiency in processing the entire CC3600 data set.

值得注意的是，GPT-4.1的價格比其前身GPT-4O低26％，這使初創企業，企業和獨立開發人員更容易使用。這種定價轉變是因為Openai面對競爭對手AI模型等競爭對手的競爭，該模型最近為處理整個CC3600數據集提供了新的基準。

With the launch of GPT-4.1, OpenAI will be phasing out GPT-4 by April 30, 2025, and deprecating the GPT-4.5 preview in its API by July 14, 2025. The company claims that GPT-4.1 delivers superior performance with lower latency, warranting the transition.

隨著GPT-4.1的推出，OpenAI將於2025年4月30日在2025年4月30日之前逐步淘汰GPT-4，並在2025年7月14日之前將GPT-4.5預覽貶值。該公司聲稱GPT-4.1在較低的延遲中提供了出色的績效，以較低的延遲，保證過渡。

Looking ahead, OpenAI has encountered delays in releasing GPT-5, originally anticipated by May 2025. CEO Sam Altman attributed the setback to challenges in seamlessly integrating the new model with existing systems. However, OpenAI is preparing for the release of its next reasoning engines—o3 and o4 mini—which are expected to follow GPT-4.1 shortly.

展望未來，Openai在發布GPT-5時遇到了延遲，最初是2025年5月。首席執行官Sam Altman將挫折歸因於將新模型與現有系統無縫集成的挑戰。但是，OpenAI正在準備發布其下一個推理引擎（O3和O4 Mini）的發布，預計將在不久後遵循GPT-4.1。