DeepSeek announced DeepSeek-V4.1-Flash-1017 on October 17, 2026, together with Search and Fetch APIs that are now open to all users.
What to know
- The model: DeepSeek calls V4.1-Flash-1017 a “Pareto improvement in performance and cost” over the previous version, with stronger benchmark results, faster task completion and shorter outputs. It is live on the website, app and API, and DeepSeek says it is open-weight on Hugging Face and ModelScope. In the API,
deepseek-flashnow points to it. These are DeepSeek’s own claims, not independent tests. - Agent tooling: the model is optimised for DeepSeek Harness (now in developer preview), Tencent CodeBuddy and WorkBuddy. DeepSeek recommends DeepSeek Harness v0.2.1-rc.1 or later.
- Search and Fetch APIs: powered by DeepSeek’s in-house search engine, priced at $0.75 per 1,000 calls for each endpoint.
- Free period: DeepSeek says Search and Fetch calls made through server-side tools in DeepSeek Harness, Claude Code and Codex are free for API users for now, with billing starting October 31, 2026 at 16:00 UTC.
Why it matters
Giving agents a cheap web search and page fetch tool is what makes them useful for research. If you build agents, see our guide to writing an agent skill, and read about the earlier DeepSeek V4.1-Flash release. More on DeepSeek and the model tracker.
Source: DeepSeek API docs. We summarise in our own words; check the source for the full details.