DeepSeek Releases V4.1 Flash Update, Opens Self-Developed Search API

DeepSeek has announced the update of its V4.1 Flash model to version V4.1-Flash-1017, with official website, app, and API model changes. The new model has shown improvements in reasoning, programming, and visual benchmarks, with a DeepSWE v1.1 benchmark score of 77.0, surpassing Claude Opus 5. The update also opens a self-developed search API, with Search and Fetch interfaces available to the public.

The update is seen as a significant step forward for DeepSeek, with its architecture innovation continuing to reduce unit costs for AI. The model’s improved value for money is expected to drive larger-scale reasoning calls, with domestic computing power poised to benefit from both industry and capital catalysts.

The update is part of a broader trend of domestic AI models gaining traction, with companies like DeepSee and others completing Day 0 adaptations and receiving attention from securities firms like Zhongyin International.