DeepSeek ships V4.1 Flash: 552B MoE on new architecture, MIT-licensed

deepseek-ai · hf · 2026-09-10

DeepSeek released DeepSeek-V4.1-Flash on Hugging Face: a 552B MoE model with a new causal-encoder-decoder architecture, native multimodal input, MIT license, fp8/8-bit weights, endpoints compatible.

Related event: DeepSeek Unveils Open-Source V4.1-Flash MoE Model(28 posts)→

Original post →

More from Infra

Infra channel →