内容来源：https://aibusiness.com/generative-ai/openai-intros-gpt-5-1-in-chatgpt

内容总结：

谷歌云赞助内容指出，企业引入生成式AI时，应优先选择能优化人机信息交互体验的场景作为切入点。

本周三，OpenAI正式在ChatGPT平台推出GPT-5.1模型。此次更新距上一代GPT-5发布仅四个月，恰逢公益组织“公民公众”因多模态模型Sora 2存在被滥用于制作深度伪造内容的风险，要求其下架该应用。

新一代模型具备三大技术突破：GPT-5.1 Instant通过自适应推理实现“思考后应答”，呈现更自然的对话温度；指令跟随能力显著提升，确保回答更精准可靠；GPT-5.1 Thinking组件可智能分配算力，对简单指令快速响应，对复杂任务加强思考深度。该模型将分阶段推送，GPT-5 Pro用户即将获得GPT-5.1 Pro升级，而内置的GPT-5.1 Auto持续保持智能路由提问至最优模型的能力。

值得注意的是，ChatGPT个性化设置迎来重要调整：原有“倾听者/机器人”模式被“默认、友好、高效”取代，新增“专业、坦率、诙谐”风格选项，用户可自定义回答的简洁度、亲和度与可读性。

高德纳分析师阿伦·钱德拉塞卡指出，本次升级彰显OpenAI提升用户体验的核心战略：通过增强对话自然度与指令执行力，最终实现模型选择的全程自动化。在GPT-5市场评价分化的背景下，此次迭代既是对产品短板的完善，也向业界展示了企业持续创新的技术节奏。专家强调，只要自动化流程足够顺畅，该演进方向将有效推动技术普惠。

中文翻译：

由谷歌云赞助
选择你的首个生成式AI应用场景
要入门生成式AI，首先应关注能够提升人类信息交互体验的领域。

此次AI模型更新展现了厂商对模型的改进成果，以及其在ChatGPT中实现自动化的努力。
OpenAI于周三在ChatGPT中推出了GPT-5.1版本。
这次更新距该生成式AI厂商发布GPT-5仅四个月，恰逢监督机构"公共公民"要求其下架Sora 2应用的后一天——该多模态模型应用因易被用于制作有害深度伪造内容而引发争议。

新版GPT-5.1搭载即时响应功能，OpenAI称其比GPT-5更具温度感且对话更自然。5.1版本还优化了指令跟随能力，使问答响应更可靠。该即时功能可运用自适应推理机制，自主判断应答前的思考时机。
同步推出的GPT-5.1思考模块能动态调整模型思考任务——厂商表示这意味着简单指令处理更迅捷，复杂任务则投入更多运算时间。

OpenAI计划未来数周逐步推送GPT-5.1，即将为GPT-5专业版用户升级至GPT-5.1专业版。与此同时，ChatGPT内置的GPT-5.1自动模式将持续将查询请求智能分配至最适宜的OpenAI模型进行处理。

厂商还将ChatGPT的定制选项从"倾听者/机器人"更名为"默认/友好/高效"，新增"专业/坦诚/奇趣"模式。用户现可通过个性化设置调节对话特性，自主定义回复的简洁度、温度感及可读性。

行业趋势
高德纳分析师阿伦·钱德拉塞卡指出，此次更新彰显了OpenAI为优化AI聊天机器人用户体验所做的努力。
"他们致力于让模型更具对话性，更精准遵循指令，最终目标是实现模型选择的完全自动化。"钱德拉塞卡表示，"这本质上契合了OpenAI推动ChatGPT内置模型选择自动化的核心目标。"

他补充道，只要自动化流程足够顺畅便不会产生显著问题。此次更新也反映出初代GPT-5获得褒贬不一的评价后，OpenAI正持续探寻模型优化空间。
"这同时向市场传递了企业创新节奏的信号。"钱德拉塞卡强调。

您可能还喜欢

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The updated AI model showcases the vendor's improvements to the model and its efforts to automate it within ChatGPT.
OpenAI introduced GPT-5.1 in ChatGPT on Wednesday.
The update comes four months after the generative AI vendor released GPT-5, and a day after watchdog group Public Citizen asked the vendor to remove its Sora 2 app from the market because of the multi-modal model app's susceptibility to being used to create harmful deepfakes.
The model upgrade comes with GPT-5.1 Instant, which OpenAI said is warmer and conversational than GPT-5. The 5.1 version also boasts improved instruction following, enabling it to answer questions more reliably. The Instant feature can use adaptive reasoning to decide when to think before responding, OpenAI said.
GPT-5.1 Thinking, another component of 5.1, can now adapt the model's thinking tasks, meaning it spends less time on easy prompts and functions and more time on complex tasks, the vendor said.
OpenAI plans to roll out GPT-5.1 gradually over the next few weeks, with an upgrade to GPT-5.1 Pro for GPT-5 Pro users soon. Meanwhile, GPT-5.1 Auto within ChatGPT will continue to route prompts or questions to the best OpenAI model that can answer them.
The vendor also renamed its customization options in ChatGPT to Default, Friendly and Efficient from Listener and Robot. It also added Professional, Candid and Quirky. Users can now tune ChatGPT's characteristics with personalization settings, such as how concise, warm, or scannable they want its responses to be.
A trend
The update demonstrates how OpenAI is striving to create a smoother user experience for its AI chatbot, according to Arun Chandrasekaran, an analyst at Gartner.
"They want these models to be more conversational, more instruction following to the point where they hope that the entire experience of selecting the right model is completely automated," Chandrasekaran said. "They want to automate the model selection within ChatGPT. That's fundamentally OpenAI's goal."
He added that as long as OpenAI makes the model selection automation process smooth for users, then there shouldn't be any significant problems with it. However, this update is also an example of how OpenAI is seeking room to improve its models after the original GPT-5 got mixed reviews.
“This is also a signal to the market around the pace of innovation within the company,” Chandrasekaran said.
You May Also Like