OpenAI发布 ChatGPT-5！免费用户首个推理模型，性能更强，幻觉更少

Core Viewpoint - OpenAI has launched its latest AI model, GPT-5, which is smarter, faster, and more useful, particularly in writing, programming, and healthcare [1][2]. Group 1: Model Features and Improvements - GPT-5 has a lower hallucination rate, meaning it fabricates answers less frequently compared to previous models [2]. - The model underwent extensive safety evaluations, including 5,000 hours of testing, and employs "safe completions" to provide advanced responses within safety limits [2][3]. - GPT-5 is designed to recognize when tasks cannot be completed, reducing guesswork and providing clearer explanations of its limitations [2][3]. Group 2: User Accessibility and Versions - GPT-5 will be available to free, Plus, Pro, and team users starting August 6, with free users having access to a reasoning model for the first time [3][4]. - OpenAI has released three different versions of GPT-5 (gpt-5, gpt-5-mini, and gpt-5-nano) tailored for varying cost and latency needs [4]. Group 3: Integration and Partnerships - Microsoft will integrate GPT-5 into its products, including Microsoft 365 Copilot and Azure AI Foundry, enhancing its capabilities for developers [4]. - Box CEO noted that GPT-5 represents a significant breakthrough, capable of retaining more information and utilizing higher-level reasoning and logic [4]. Group 4: User Experience - The interaction with GPT-5 feels more natural and human-like, with the model likened to having a team of PhD-level experts available at all times [5].