GPT-4.1和GPT-4.1 MINI现已在Chatgpt中找到

AI型号并不总是需要开创性的才能有所作为。有时，速度，结构和聚焦的安静变化会影响我们的工作方式

GPT-4.1 and GPT-4.1 mini, previously released by OpenAI in April via its API, are now rolling out to ChatGPT users, marking a silent yet significant shift in the chat service's capabilities.

GPT-4.1和GPT-4.1 Mini先前于4月通过其API发行，现在正在向ChatGpt用户推出，这标志着聊天服务能力的无声而又重大的转变。

The models, which drew attention for their arrival without the usual technical or safety report from OpenAI, are designed for performance, particularly in coding and instruction-following, areas where OpenAI claims GPT-4.1 shows "huge gains."

这些模型在没有OpenAI的常规技术或安全报告的情况下引起了人们的注意，旨在绩效，尤其是在编码和指导跟踪方面，OpenAI索赔GPT-4.1显示“巨大的收益”。

Now, reaching a wider user base, GPT-4.1 arrives to Plus, Team, and Enterprise subscribers, while GPT-4.1 mini will take the place of GPT-4.0 mini as the fallback for both free and paid tiers. This brings a more powerful and efficient text model into daily use, with improvements in consistency, speed, and prompt handling to be expected.

现在，到达更广泛的用户群，GPT-4.1到达Plus，Team和Enterprise订阅者，而GPT-4.1 Mini将代替GPT-4.0 Mini作为免费和付费层的后备。这将更强大，更有效的文本模型用于日常使用，并预期一致性，速度和及时处理。

GPT-4.1 is also noted for its 1 million-token context window, nearly eight times that of GPT-4-turbo (often referred to informally as GPT-4.5), which is described by OpenAI as having 128,000 tokens. This allows for processing larger documents, longer conversations, and tasks like codebase analysis with fewer interruptions.

GPT-4.1还以其100万个to录窗口而闻名，几乎是GPT-4-Turbo（通常被非正式地称为GPT-4.5）的八倍，Openai将其描述为具有128,000个令牌。这允许处理较大的文档，更长的对话和诸如代码库分析之类的任务，而中断较少。

While early benchmarks like 54.6% on the SWE-bench Verified for GPT-4.1 in its strengths are noted, and users may notice trade-offs in tone or conversational nuance compared to previous versions. As part of this, GPT-4.0 mini has been retired, and OpenAI plans to fully phase out GPT-4-turbo by July 2025, from its API, to simplify its model lineup.

虽然早期的基准标有54.6％的SWE Bench，其优势已验证了GPT-4.1的优势，并且与以前的版本相比，用户可能会注意到语调或对话差异的权衡。为此，GPT-4.0 Mini已退休，OpenAI计划在2025年7月之前从其API中完全淘汰GPT-4-Turbo，以简化其模型阵容。