DeepSeek ships V4.1 Flash snapshot 1017
The deepseek-flash API now serves DeepSeek-V4.1-Flash-1017. Search and Fetch APIs are open, at $0.75 per 1,000 calls.
Official source (opens in a new tab)
DeepSeek's news post for the snapshot DeepSeek-V4.1-Flash-1017 says it is live on the website, the app, and the API. DeepSeek says it has stronger benchmark performance, finishes tasks faster, and writes shorter outputs than the previous version, which they call a Pareto improvement in performance and cost. The post says it is optimized for DeepSeek Harness, Tencent CodeBuddy, and WorkBuddy, and that DeepSeek Harness v0.2.1-rc.1 or later is the version to use. The API docs say the model name deepseek-flash is now served by this snapshot. Weights are on Hugging Face and ModelScope.
The same post opens Search and Fetch APIs to everyone, at $0.75 per 1,000 calls. Calls made through server-side tools in DeepSeek Harness, Claude Code, and Codex are free for DeepSeek API users until 31 October 2026 at 16:00 UTC.
Who it's for: people already calling DeepSeek's flash model, or teams that want the weights. It is a snapshot update, not a new chat product on this site's lists.
Fits: Best for Vibe Coding only as a model you might call from an agent. It does not change which coding tool ranks first.