DeepSeek updates V4.1 Flash model to speed up task execution

Chinese artificial intelligence startup Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. has released an updated version of its V4.1-Flash model, SiliconANGLE AI and trendingtopics.eu reported. The updated model checkpoint, named DeepSeek-V4.1-Flash-1017, was released on October 17, 2026. The release arrives just over five weeks after the original V4.1-Flash launched as the smallest model in a new architecture family. [1][3][4][8]

The company describes the release on its Hugging Face model card as a minor update that reduces task execution time and output length, according to SiliconANGLE AI. DeepSeek did not disclose the exact magnitude of the speed improvements. MarkTechPost reported that the updated model scored 77.0 on the DeepSWE v1.1 benchmark. [1][2][8]

DeepSeek operates out of Hangzhou, Zhejiang, and is owned and funded by the Chinese hedge fund High-Flyer. The company develops open weights large language models. The updated model release was accompanied by the launch of dedicated Search and Fetch application programming interfaces, as reported by trendingtopics.eu, Dealroom, and FTMQ SI. [4][6][7][8]

As FTMQ SI reported earlier, the company set the pricing for its Search and Fetch application programming interfaces at 0.75 dollars per 1,000 calls. The Decoder reported that the update aims to lower agentic execution costs while competing directly against larger model alternatives. [5][6][8]