DeepSeek has released an updated iteration of its model, designated as DeepSeek-V4.1-Flash-1017, claiming significant gains in speed and efficiency. The Hangzhou-based startup stated that the minor update delivers stronger benchmark performance alongside shorter responses and faster task completion times. According to technical documentation published on Hugging Face, the new model introduces substantial improvements across reasoning, coding, visual benchmarks, and agent-based workflows. The release is immediately available to users and developers across DeepSeek’s website, mobile app, and developer platform. The rapid deployment intensifies the ongoing competition between Chinese open-weight models and proprietary Western systems from companies like OpenAI and Anthropic.

Full report : The Chinese AI startup rolls out a minor update targeting improved reasoning, coding, and agentic task execution to intensify competition with U.S. and domestic rivals.