On October 17, DeepSeek released V4.1-Flash-1017, a minor update to the V4.1-Flash model launched in September — yet it delivers notable gains across many benchmarks.
First, coding and agentic ability: DeepSWE v1.1 rose from 74.2 to 77.0, ahead of Claude Opus 5 at 74.0; Terminal-Bench 4.0 jumped from 31.2 to 42.3; Automation-Bench climbed from 54.8 to 60.5. DeepSeek says all 11 benchmarks improved, with faster task completion and shorter outputs.
Second, an open ecosystem: the model weights are open-sourced on Hugging Face under the MIT license, free for developers to use. DeepSeek also opened its Search & Fetch APIs to outside developers for the first time, with some calls free until November 1, giving agent applications a first-party web retrieval layer.
The update is another reminder that open models are squeezing closed models' price premium with a "cheap and capable" combination, giving developers a more cost-effective frontier option.
Source: Read the original article. This is a summarized digest for reference only.
