MyPrivateClaw

DeepSeek officially released the DeepSeek-V4.1-Flash model on 2026-09-10 | Model Release

V4.1 Flash is live on the DeepSeek API under the model name deepseek flash, while the previous generation V4 Flash and V4 Flash Vision Exp models were retired…

Published on MyPrivateClaw

Sep 10, 2026, 5:15 PM UTC

Coverage date

Sep 10, 2026

Last updated

Sep 10, 2026, 5:15 PM UTC

News summary

DeepSeek V4.1 Flash is a 552B parameter MoE model using a new Causal Encoder Decoder architecture with 8B active parameters for input and 16B for output. On the DeepSeek benchmark table, DeepSeek V4.1 Flash scores GPQA Diamond 90.9, HLE 36.8, Codeforces rating 3471, and Terminal Bench 2.1 90.6. DeepSeek pointed to an open source Hugging Face model page and technical report for DeepSeek V4.1 Flash, while also stating it would explore open source inference support and larger scale deployment options.