Latest / Frontier models and safety

Model releaseFrontierChinaConfirmed

DeepSeek releases V4.1-Flash-1017 update with open weights

See how this story connects

DeepSeek released DeepSeek-V4.1-Flash-1017 on 17 October 2026, an update to the V4.1-Flash model of 10 September, with open weights under the MIT licence on Hugging Face and ModelScope and live on its website, app and API. DeepSeek says the update scores higher than V4.1-Flash on reasoning, coding, agentic and visual benchmarks and finishes tasks with shorter outputs. Its own table puts it at 77.0 on DeepSWE v1.1 against 73.0 for GPT-5.6 Sol and 74.0 for Claude Opus 5, and at 42.3 on Terminal-Bench 4.0 against 51.8 for Opus 5. The same release opened DeepSeek’s Search and Fetch APIs to all users at US$0.75 per 1,000 calls, with billing from 31 October 2026.

Why it matters

An open-weights model that DeepSeek’s own tests put ahead of OpenAI’s and Anthropic’s models on some coding and agent benchmarks can be downloaded and run under the MIT licence.

When
17 Oct 2026
ID
WTA-N-2026-0324
Where
China
Who
DeepSeek

Summary by wheresthe.ai, from DeepSeek API docs and the reporting credited above. Checked against the source on 19 Oct 2026.

Source copy: download the page as saved on 19 Oct 2026 (PDF). Also held by the Internet Archive.

Next stop on this Storyline: Led toDeepSeek open-sources DeepInfer, the inference framework behind its production services

Cite this story

WTA-N-2026-0324: “DeepSeek releases V4.1-Flash-1017 update with open weights.” wheresthe.ai, 17 Oct 2026. https://wheresthe.ai/d/cn-deepseek-v4-1-flash-1017-2026-10/ (source: DeepSeek API docs: V4.1-Flash-1017).

Spotted an error? Email [email protected].