China’s DeepSeek has unveiled what it calls a faster and leaner version of its artificial intelligence model.

Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

The company describes the DeepSeek-V4.1-Flash-1017 model, announced Saturday (Oct. 17) as offering “[s]tronger benchmark performance, faster task completion, and shorter outputs than the previous version.”

DeepSeek added that this model employs “reinforcement learning with self-feedback” to bolster end-to-end project development for things like game creation to 3D modeling.

The new open-source model provides “substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks,” the startup wrote in its technical report on the AI platform HuggingFace.

A report on the launch by Bloomberg News pointed out that DeepSeek recently suffered a five-hour outage that left its servers unreachable. Some experts have attributed the incident to insufficient computing resources, the report said.

Please add us to your preferred sources list so our news, data and interviews show up in your feed. Thanks!

Bloomberg added that slowdowns and service disruptions have risen recently for DeepSeek, whose lower-cost AI models are popular among developers, putting strain on its services. The report also noted that this launch highlights the increasing competition between China’s open-weight AI models and the proprietary versions offered by their U.S. contemporaries, which are leading performers but cost much more.

PYMNTS wrote earlier this year about the open-source debate from the point of view of chief financial officers of middle market companies, who are dealing with a question more financial than ideological : not whether open-weight AI will defeat proprietary AI but whether the savings, flexibility and control provided by open models are enough to justify taking on greater responsibility for the infrastructure beneath them.

Source link