Its model card describes the release as a minor update intended to reduce task execution time and output length. The company did not state the size of either reduction.
According to the report, the update resolved 77% of tasks on DeepSeek’s DeepSWE v1.1 software-engineering test, compared with 74.2% for the September version. API users access it through the existing deepseek-flash model ID at unchanged prices. DeepSeek recommends Harness version 0.2.1-rc.1 or later, and has released the weights under the MIT license.
DeepSeek also opened a Search API, returning up to 50 ranked web results per query, and a Fetch API that converts one web page to Markdown. Each costs $0.75 per 1,000 successful requests; calls through DeepSeek Harness, Claude Code, or Codex are free until November 1. The report does not independently establish the claimed speed improvement, benchmark result, or real-world quality of the search tools.