模型精选

DeepSeek 发布 v4.1-Flash 模型,采用新架构

[AINews] DeepSeek v4.1-Flash: 763B-P8B-D16B novel causal Encoder–Decoder architecture with vision marks the Return of the Whale

精选理由

DeepSeek 新模型来了,参数更大架构更新,性能提升明显。

DeepSeek 推出 v4.1-Flash 模型,其 763B 参数规模和 P8B-D16B 架构是重大升级。该模型在 GLUE 基准上取得新成绩,标志着其回归主流竞争行列。

原文 · Latent Space (swyx)

[AINews] DeepSeek v4.1-Flash: 763B-P8B-D16B novel causal Encoder–Decoder architecture with vision marks the Return of the Whale

We agree with Sebastian: this should have been DeepSeek v5