News
OpenAI launches GPT-5
2 min read
Source: Telegram AI频道
OpenAI launches the GPT-5.6 model: a new "Think" button and a calculation intensity slider are added to improve the accuracy and simplicity of answers. OpenAI recently released its latest version of the AI model GPT-5.6. This update aims to solve the problems that users have long reported, such as long language, too much nonsense, and "pretending to understand". It strives to provide more refined and accurate answers and reduce the probability of giving wrong facts (i.e., AI illusions). In this update, the GPT-5.6 Sol version brings a number of core improvements. Officials said the new model can provide more targeted answers and can automatically adjust the level of detail of answers based on specific questions. At the same time, it reduces unnecessary formatting, and when it discovers factual errors in user questions, it will proactively provide reasonable corrections instead of blindly catering to or obeying users. In addition, the most eye-catching change in the new interface is the introduction of a "think" button and functions similar to the "calculation intensity slider". Users can switch between "deep thinking mode" and "immediate response mode", or adjust the model's computing resource investment through the slider. The lower the slider is set, the faster the response and the less computing resources are consumed; while raising the slider means investing more computing power to obtain more detailed and in-depth analysis results. This move is believed to help guide users to reasonably allocate computing power according to the difficulty of the problem, thereby reducing the company's operating costs. For free users, OpenAI will provide the GPT-5.6 Luna version by default, and also gives users the authority to use the "think" button to deal with complex problems. Although OpenAI has been continuously improving the factual reliability of its models, relevant industry analysts still remind that artificial intelligence models may still produce fictitious content in nature. When it comes to professional fields such as law, medical care, and finance, users must not rely too much on AI-generated text as the final work product. via cnBeta.COM - Chinese industry information station (author: source: cnBeta.COM)