IT之家 9 月 30 日消息，美国 AI 基础设施公司 Baseten 今天宣布与 OpenAI 达成合作，成为 OpenAI B2B 市面首批开源模型推理服务给出商之一。企业使用者可借助 Codex 使用 Kimi K3 和 GLM 5.3 等开源模型，意味着中国开源模型首次进入 OpenAI 企业采购体系。
IT之家了解到，Kimi K3 是月之暗面开发的大模型。官方宣称其是首个参数规模达 2.8 万亿的开源模型。该模型原生兼容视觉能力，具备 100 万 Token 的上下文窗口，格外合适执行长时间编程、多文档分析等工作。
如今，OpenAI 企业使用者可借助 Codex 或 Responses API 调用 Kimi K3 等开源模型，以相同的预算取得更多 AI 能力。Baseten 承诺其给出的所有模型均运行在美国服务器，针对所有提示词实行零数据保留（ZDR）。
本文来源：https://www.ithome.com/1/008/864.htm
若有版权问题，请来信告知：13951431913#139.com 或在右方提交评论。
Thanks to IT Home user "Buyao Ya" for submitting the lead!
According to IT Home on September 30, US-based AI infrastructure company Baseten today announced a partnership with OpenAI, becoming one of the first commercial providers of open-source model inference services for OpenAI's B2B market. Enterprise users can utilize Kimi K3 and GLM 5.3, among other open-source models, through Codex, marking the first time Chinese open-source models have been integrated into OpenAI's enterprise procurement system.
IT Home learned that Kimi K3 is a large model developed by Moonshot AI. The company officially claims it is the first open-source model with a parameter scale of 2.8 trillion. This model is natively compatible with vision capabilities and features a 1-million-token context window, making it particularly suitable for tasks such as long-duration coding and multi-document analysis.
Now, OpenAI enterprise users can call open-source models like Kimi K3 via Codex or the Responses API to access more AI capabilities with the same budget. Baseten has committed that all models it offers run on US servers and implements zero data retention (ZDR) for all prompts.