模型官方一手精选

DeepSeek 发布 V4.1 Flash,V4 Pro 于 9 月 14 日停服

DeepSeek Ships V4.1 Flash GA With Causal-Encoder-Decoder MoE as V4 Pro Retires

精选理由

DeepSeek 新发布的 V4.1 Flash 模型,规模更大,架构更先进,价格也调整了,原 V4 Pro 在 9 月 14 日就停用了,想试试的话可以看看。

DeepSeek 推出 V4.1 Flash 模型,采用因果编码器-解码器混合专家架构,规模达 552B,并内置原生视觉能力。新版本定价自 9 月 10 日起生效,原 V4 Pro 服务将于 9 月 14 日转向 Flash,应用模式合并为智能模式。

原文 · pandaily

DeepSeek Ships V4.1 Flash GA With Causal-Encoder-Decoder MoE as V4 Pro Retires

DeepSeek has released V4.1 Flash for general availability, a 552B MoE model with Causal-Encoder-Decoder architecture and native vision. New Flash pricing started September 10, V4 Pro routes to Flash on September 14, and the app merges modes into one smart mode.