Chinese artificial intelligence startup Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co. Ltd. has released an updated version of its V4.1 Flash model that claims to complete tasks faster.
DeepSeek-V4.1-Flash-1017 arrives just over five weeks after the company launched the original V4.1 Flash as the smallest model in a new architectural family. Hugging Face’s model card describes the release as a “small update” that “reduces task execution time and output length.” DeepSeek didn’t say by how much.
Price was the selling point in September, with off-peak application programming interface prices starting at 15 cents per million uncached input tokens. Following the release, Bloomberg reported on Saturday, shares of Chinese rivals MiniMax Group Inc. and Alibaba Group Holding Ltd. fell. in Hong Kong.
The update solved 77% of the problems on the DeepSWE v1.1 software engineering test in DeepSeek’s own testing, up from 74.2% in the September version. Claude Opus 5 from Anthropic PBC is in the same chart with 74%. Seven of the 11 benchmarks published by DeepSeek are still based on Anthropic’s model, and the lead on Humanity’s Last Exam is almost 14 points.
DeepSeek said it optimized the update for agent software, including the DeepSeek Harness tool and two Tencent Holdings Ltd. products, coding assistant CodeBuddy and WorkBuddy, a desktop agent for office work. WorkBuddy has been offering V4.1 Flash since DeepSeek released the model on September 10th.
Agent Builders have received a second release from DeepSeek. The company opened its internal search engine to external developers via 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 via DeepSeek Harness, Anthropic’s Claude Code or OpenAI Group PBC’s Codex are free until November 1st.
The releases follow a nearly five-hour API outage on October 15, which Yicai Global ranked among DeepSeek’s worst this year. The company subsequently apologized and users received credits tied to their spending the week before. DeepSeek also said it would remove individual dependency points in its systems to prevent the bug from recurring. Some experts quoted by Bloomberg blamed a lack of computing capacity.
The updated model is now available on DeepSeek’s web and mobile apps. API customers can access it via the existing Deepseek Flash model ID at unchanged pricing, with the changelog recommending Harness version 0.2.1-rc.1 or later. The weights are available on Hugging Face under the MIT license.
Image: DeepSeek
Support our mission to keep content open and free by interacting with theCUBE community. Join theCUBE Alumni Trust Networkwhere technology leaders connect, share information and create opportunities.
- Over 15 million viewers of theCUBE videosto spark conversations about AI, cloud, cybersecurity and more
- Over 11.4k theCUBE alumni — Connect with more than 11,400 technology and business leaders shaping the future through a unique, trusted network
Founded by tech visionaries John Furrier and Dave Vellante, SiliconANGLE Media has built a dynamic ecosystem of industry-leading digital media brands, reaching over 15 million elite tech professionals. Our new proprietary theCUBE AI Video Cloud is a game-changer in audience engagement, leveraging the theCUBEai.com neural network to help technology companies make data-driven decisions and stay at the forefront of the industry conversation.
