DeepSeek unveils updated AI model claimed to be faster than its predecessor

DeepSeek on Saturday unveiled an updated version of its AI model, V4.1-Flash-1017. According to the company, it runs faster, produces shorter responses, and performs better on benchmarks than the previous version. The release intensifies the Chinese start-up’s competition with U.S. giants Anthropic and OpenAI, as well as domestic AI labs Moonshot Kimi and Z.ai, Bloomberg reports.

DeepSeek unveiled the model in a series of posts on X. According to its developers, the update delivers “stronger benchmark performance, faster task completion, and shorter outputs than the previous version.” It is already available through DeepSeek’s web interface, mobile app, and developer platform.

In a technical report on Hugging Face, the start-up reported “substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks.” The new version expands the model’s capabilities across tasks ranging from game creation to 3D modeling. The company describes it as a minor update to the previous V4.1-Flash.

DeepSeek also disclosed results on Humanity’s Last Exam, a test designed to assess AI capabilities through thousands of questions in math, sciences, and humanities. The updated model scored 36.8-42.6, versus 36.8-39.1 for the previous version.

The release came amid technical problems at DeepSeek. Earlier this week, the start-up’s servers were unreachable for around five hours. Some experts attributed the outage to insufficient computing capacity, Bloomberg notes. Service disruptions have become more frequent in recent months as developers’ demand for the company’s extremely cheap models has grown.

The preceding model, V4.1-Flash, was released in September at such a steep discount that shares of DeepSeek’s rivals MiniMax Group, Z.ai, and even giant Alibaba plunged on the Hong Kong stock exchange. DeepSeek was already known as one of the AI market’s biggest disruptors. In January 2025, the release of its R1 model, which the company claimed had been developed for less than $6 million, triggered a $1 trillion selloff in U.S. and European tech stocks in a single day, including a 15% drop in Nvidia.

In August, however, DeepSeek sharply raised prices for its flagship V4 models. Bloomberg linked the move to the company’s preparations for an IPO and efforts to improve profitability. At the time, the company was also in talks over a funding round at a valuation of up to $71 billion.