News · DeepSeek · 2 min read
Event date: · Published:
DeepSeek upgrades deepseek-flash to DeepSeek-V4.1-Flash-1017 and reports higher coding and agent scores
On October 17, 2026, DeepSeek upgraded its web and app versions to DeepSeek-V4.1-Flash-1017, served behind the deepseek-flash API name at unchanged pricing, and published MIT-licensed weights on Hugging Face and ModelScope.
Archive entry: this covers an event from October 17, 2026, and we published it on November 8, 2026. The details come from sources dated at the time and may have changed since.
On October 17, 2026, DeepSeek upgraded its web and app versions to DeepSeek-V4.1-Flash-1017. The model was served behind the deepseek-flash API name, with the API call method and pricing unchanged, as described in the DeepSeek API Docs announcement.
What changed in the model
- DeepSeek reported coding and agent scores of Terminal-Bench 4.0 42.3 (Claude Code framework), DeepSWE v1.1 77.0, ProgramBench 29.1, NL2Repo-Bench 68.2, Automation-Bench v1.0.6 60.5 and Agents' Last Exam 35.4.
- Other reported scores were HLE without tools 42.6, Apex Bench v1.1 72.1, Chartography 83.0, BabyVision 92.5 and ZeroBench-main 51.0.
- DeepSeek said it reduced task runtime and output length, for lower cost and higher speed.
- Reinforcement learning with self-feedback targeted building whole projects such as games and 3D scenes.
- DeepSeek optimized the model for DeepSeek Harness, Tencent CodeBuddy and WorkBuddy, and recommended DeepSeek Harness 0.2.1-rc.1 or later.
What it means for people building with AI
Anyone calling deepseek-flash got the upgraded coding model automatically, at the same price and with the same API call method. DeepSeek said the shorter task runtime and output length were meant to lower cost and raise speed, and its recommendation of DeepSeek Harness 0.2.1-rc.1 or later told builders which tooling the model was tuned for.
Availability and sources
DeepSeek published the weights on Hugging Face as deepseek-ai/DeepSeek-V4.1-Flash-1017 under the MIT license, and on ModelScope. Its Models & Pricing page named DeepSeek-V4.1-Flash-1017 as the model behind deepseek-flash, and the DeepSeek API change log records the release on October 17, 2026.
Sources
- DeepSeek-V4.1-Flash-1017: Wish-Driven Programming Made Real | DeepSeek API Docs (DeepSeek, October 17, 2026, official)
- deepseek-ai/DeepSeek-V4.1-Flash-1017 (DeepSeek, official)
- Change Log | DeepSeek API Docs (DeepSeek, October 17, 2026, official)
- Models & Pricing | DeepSeek API Docs (DeepSeek, official)
Written with AI assistance, strictly from the sources listed above. Before it went live, automated checks verified its dates, sources and every language edition.