DeepSeek releases V4.1-Flash-1017 with open weights, and opens its Search and Fetch APIs at $0.75 per 1,000 calls
DeepSeek calls the update “a Pareto improvement in performance and cost” over the version it replaces, and tunes it for its own harness and Tencent's coding assistants.
DeepSeek released DeepSeek-V4.1-Flash-1017 on 17 October. Its release note says the model gives “Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost.” The model is live on DeepSeek's website, app and API, and the weights are on Hugging Face under the MIT License.
DeepSeek says the model is tuned for DeepSeek Harness and for Tencent's CodeBuddy and WorkBuddy assistants, and that it uses reinforcement learning with self-feedback to improve end-to-end project work. The model card lists a context of 1 million tokens. In DeepSeek's own table, the new version improves on V4.1-Flash on Terminal-Bench 4.0 and on Humanity's Last Exam without tools.
The same note opens DeepSeek's Search and Fetch APIs to all users at $0.75 per 1,000 calls for each endpoint. Calls made through server-side tools in DeepSeek Harness, Claude Code and Codex are free until billing starts on 31 October. The note gives no change to the price of the model itself.
It is the second open-weight release in two days, after Reflection's Beam on 16 October, and follows Meta's Muse Spark Open, which the Switchboard reported on 16 October.
- Confirmed DeepSeek released V4.1-Flash-1017 on 17 October, live on its website, app and API. DeepSeek API Docs
- Claimed DeepSeek calls it “a Pareto improvement in performance and cost” over the previous version. DeepSeek API Docs
- Confirmed Search and Fetch APIs open to all users at $0.75 per 1,000 calls per endpoint, with billing from 31 October. DeepSeek API Docs
Models & releasesAgents in the wild
Today in the October 17, 2026 edition · front page