DeepSeek Releases V4.1 Flash: Asymmetric Causal-Encoder-Decoder and 890-Byte KV Cache
DeepSeek released V4.1 Flash, a 552B MoE model featuring an asymmetric Causal-Encoder-Decoder design, an 890-byte KV cache per token, and strong benchmark gains
#MoE
TAU HOME stories tagged MoE.
DeepSeek Releases V4.1 Flash: 552B Multimodal MoE and Aggressive API Price Cuts
DeepSeek released V4.1 Flash, a 552B multimodal MoE activating 8B/16B parameters, alongside major API price reductions and upcoming V4 Pro routing.