DeepSeek has introduced a new checkpoint of its V4.1-Flash model, designated V4.1-Flash-1017. The model is now available through the company's website, app, and API. According to DeepSeek, this version delivers stronger scores, faster task completion, and shorter outputs than the September V4.1-Flash release.
The company reports a score of 77.0 on DeepSWE, a benchmark referenced in the announcement. Alongside the model update, DeepSeek has opened a search API, expanding access to its search capabilities. The source does not provide further detail on the benchmark methodology or the API's specific features.
Because this report is based on a single source, the performance claims and feature descriptions come directly from DeepSeek's announcement and have not been independently verified. No contrasting views or additional context are available in the source material.