AI News Feed
Market watch
Xbox Series X Runs Four PS5 Games Through PS5X, Limited to Developer ModeTechRadar Publishes Hints and Answers for Quordle, Strands and Connections on October 19Ukrainian FPV drone simulator Remote Reaper reaches Steam 1.0Engadget explains modem, router and gateway differences and the case for buying your own routerDeepSeek releases V4.1-Flash-1017 update and opens its search engine to developersStarlink’s two $55 plans diverge on speed and data caps
Xbox Series X Runs Four PS5 Games Through PS5X, Limited to Developer ModeTechRadar Publishes Hints and Answers for Quordle, Strands and Connections on October 19Ukrainian FPV drone simulator Remote Reaper reaches Steam 1.0Engadget explains modem, router and gateway differences and the case for buying your own routerDeepSeek releases V4.1-Flash-1017 update and opens its search engine to developersStarlink’s two $55 plans diverge on speed and data caps
Large Language Models

DeepSeek releases V4.1-Flash-1017 update and opens its search engine to developers

DeepSeek released V4.1-Flash-1017, an update it says cuts task execution time and output length, and opened its search engine to outside developers.

AI News Feed · 2026-10-19 08:41

The release arrived a little over five weeks after DeepSeek launched the original V4.1-Flash. The model card on Hugging Face describes the new version as a "minor update" that "reduces task execution time and output length"; DeepSeek did not say by how much. The company released the original model on Sept. 10, when price was the selling point: off-peak application programming interface rates started at 15 cents per million uncached input tokens. Bloomberg reported on Saturday that shares of Chinese rivals MiniMax Group and Alibaba Group Holding fell in Hong Kong after that launch.

In DeepSeek's own testing, the update resolved 77% of tasks on the DeepSWE v1.1 software engineering test, up from 74.2% for the September version. Anthropic's Claude Opus 5 sits at 74% on the same table, according to the figures published by DeepSeek, which said seven of the 11 benchmarks it released still favor Anthropic's model, with the margin on Humanity's Last Exam running to almost 14 points.

DeepSeek said it tuned the update for agent software, including its own DeepSeek Harness tool and two Tencent Holdings products, the CodeBuddy coding assistant and WorkBuddy, a desktop agent for office work. WorkBuddy has offered V4.1-Flash since the model's Sept. 10 release.

Agent developers received a second release. DeepSeek opened its internal search engine to outside developers through a Search API that returns up to 50 ranked web results per query. A separate Fetch API converts a single web page to Markdown. Both cost 75 cents per 1,000 successful requests, and calls made through DeepSeek Harness, Anthropic's Claude Code or OpenAI's Codex are free until Nov. 1.

The releases follow a nearly five-hour API outage on Oct. 15 that Yicai Global counted among DeepSeek's worst this year. The company apologized afterward and gave users credits pegged to what they had spent the previous week. DeepSeek also said it would remove single points of dependency in its systems to keep the failure from recurring. Some experts quoted by Bloomberg blamed a shortage of computing capacity.

The updated model is live in DeepSeek's web and mobile apps. API customers reach it through the existing deepseek-flash model ID at unchanged prices, and the change log recommends Harness version 0.2.1-rc.1 or later. The weights are published on Hugging Face under the MIT license.