OPenAI GPT-4.1 in the API
Here is the translation:
The OpenAI GPT-4.1 model has been released. It’s a log analysis assistant that outperforms GPT-4 by 60% in performance.
Key features of GPT-4.1 include:
- The model can accurately answer questions about log files
- Input and output costs are $2 per million tokens and $8 per million tokens respectively
- GPT-4.1 mini and nano versions have input and output costs of $0.40 and $1.60, and $0.10 and $0.40 respectively
- The cache discount for prompt words has been increased to 75%
- Long context requests are now included in standard token-based billing without additional fees
Additionally, GPT-4.1 will begin deprecating the GPT-4.5 preview version in API, with a deprecated date of July 14, 2025.
OpenAI also released a new model called o3 that can integrate concepts across disciplines. This model can suggest novel experimental approaches involving fields such as nuclear fusion to pathogen detection.
Furthermore, OpenAI is developing an advanced programming agent named A-SWE that possesses autonomous abilities and can independently complete software development processes, including QA, testing code, fixing bugs, writing documentation, etc.
Translation
OpenAI 的 GPT-4.1 模型已经发布。它是一个日志分析助手，与 GPT-4 相比，GPT-4.1 在性能方面提升了 60%。
GPT-4.1 的特点包括：
- 模型可以准确回答关于日志文件的问题
- 输入和输出的成本分别为每百万 token 2 美元和 8 美元
- GPT-4.1 mini 和 nano 版本的输入和输出成本分别为 0.40 美元和 1.60 美元，和 0.1 美元和 0.4 美元
- 提示词缓存折扣提高到了 75%
- 长上下文请求已经包含在标准的按 Token 计费内，无需额外费用
此外，GPT-4.1 将开始在 API 中弃用 GPT-4.5 预览版，弃用时间为 2025 年 7 月 14 日。
OpenAI 还推出了一个新模型 o3，它可以做到跨学科的整合概念。这个模型可以提出涉及从核聚变到病原体检测等领域的全新实验思路。
此外，OpenAI 还在开发更先进的编程 Agent 名为 A-SWE，它具有自主能力，可以独立完成软件开发过程，包括完成 QA、测试程序、修复 bug、写文档等工作。
Reference:
https://openai.com/index/gpt-4-1/