📄 中文摘要
OpenAI 于 2026 年 3 月 3 日发布了最新的对话模型更新——GPT-5.3 Instant，旨在解决用户在日常使用 ChatGPT 时遇到的诸多问题，如不必要的拒绝、说教式的免责声明、陈旧的网络答案以及有时像合规官而非助手的语气。GPT-5.3 Instant 是迄今为止 OpenAI 最精致的对话模型，减少了多达 26.8% 的幻觉现象，消除了大多数不必要的拒绝，能够综合网络结果而不是简单列出链接，并且在表达上展现出更大的范围和具体性。
OpenAI GPT-5.3 Instant：减少拒绝、更好的网络答案和更流畅的 ChatGPT
📄 English Summary
OpenAI GPT-5.3 Instant: Fewer Refusals, Better Web Answers, and a Smoother ChatGPT
OpenAI released its most user-visible model update on March 3, 2026, named GPT-5.3 Instant, focusing on addressing common frustrations users face with ChatGPT. This update aims to reduce unnecessary refusals, preachy disclaimers, stale web answers, and a tone that sometimes resembles that of a compliance officer rather than a helpful assistant. GPT-5.3 Instant is OpenAI's most polished conversational model to date, achieving a reduction in hallucinations by up to 26.8%, eliminating most unnecessary refusals, synthesizing web results instead of merely providing link lists, and demonstrating noticeably greater range and specificity in its responses.