DeepSeek has released a new version of its artificial intelligence model that the Chinese company says is faster, leaner, and better at completing work. The open-weight model, called DeepSeek-V4.1-Flash-1017, was announced on October 17 and is available through DeepSeek’s website, app, and developer platform.
According to the company, the update delivers stronger benchmark performance, faster task completion, and shorter outputs than the previous version. DeepSeek says the model was trained with reinforcement learning with self-feedback, a technique aimed at improving end-to-end project work such as game development and 3D modeling.
The company claims substantial gains across reasoning, coding, agentic tasks, and visual benchmarks in a technical report posted on Hugging Face. The release keeps pressure on closed models such as ChatGPT and Gemini, which cost significantly more to run, while giving developers a free model they can download and customize themselves — a key reason open models like DeepSeek stay popular with builders.
The launch arrives as DeepSeek reportedly nears a $12 billion funding round ahead of a planned 2027 IPO, and follows recent service disruptions the company has attributed to limited computing capacity.
Source: PYMNTS
