月之暗面最强 AI 模型:Kimi K3.1 前端标识泄露,预计将在近期发布Moonshot AI's Strongest Model: Kimi K3.1 Frontend Badge Leaks, Expected Soon


AI free token/AI免费积分
感谢IT之家网友 一只Zenon 的线索投递!

IT之家 9 月 28 日消息,今日有多名开发者注意到月之暗面 API 平台的后台模型注册表中浮现了“kimi-k3-1”标识,该模型标识借助了接口探测并可调用。

月之暗面最强 AI 模型:Kimi K3.1 前端标识泄露,预计将在近期发布

今日晚些时候,Kimi 官方开放平台也浮现了 K3.1 或 K3 11111 模型的预告,兼容 1M tokens 上下文窗口。

月之暗面最强 AI 模型:Kimi K3.1 前端标识泄露,预计将在近期发布

Kimi K3.1 预计兼容最高 100 万 Token 上下文窗口,并给出 Low、High、Max 三档推理强度选项,可能引入 Agent 智能体模式、Swarm 多智能体协作以及搜索和批处理等任务模式。有开发者分析认为,多档推理强度的外观可能对应不同的算力资源规格与计费模型。

Kimi 开放平台表明,Kimi K3.1 的 API 售价与现有 K3 相同,但也不排除是占位符的可能。

相关阅读:

本文来源:https://www.ithome.com/1/008/033.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之家 netizen 一只Zenon for the tip!

IT之家, September 28 - Today, several developers noticed that the identifier "kimi-k3-1" appeared in the backend model registry of the Moonshot AI API platform. The model identifier was discovered through API probing and can be invoked.

Moonshot AI's Strongest Model: Kimi K3.1 Frontend Badge Leaks, Expected Soon

Later today, the Kimi official open platform also revealed a preview of the K3.1 or K3 11111 model, compatible with a 1M tokens context window.

Moonshot AI's Strongest Model: Kimi K3.1 Frontend Badge Leaks, Expected Soon

Kimi K3.1 is expected to support a context window of up to 1 million tokens, offering three reasoning intensity options: Low, High, and Max. It may introduce Agent mode, Swarm multi-agent collaboration, as well as task modes such as search and batch processing. Some developers believe that the presence of multiple reasoning intensity levels may correspond to different computing resource specifications and billing models.

The Kimi open platform states that the API price of Kimi K3.1 is the same as the existing K3, but the possibility that it is a placeholder cannot be ruled out.

Related reading:

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry