DeepSeek Unveils Upgraded AI Model, Claims It Outperforms OpenAI and Anthropic at Lower Cost

DeepSeek, a Chinese artificial intelligence startup, has started rolling out an upgraded version of an AI model it first introduced in September, arguing that the new release delivers stronger performance than leading U.S. systems while remaining significantly cheaper to run.

The company says the updated model outperforms products from American rivals including OpenAI and Anthropic on certain benchmarks and evaluation measures, a claim likely to draw attention as global investors and technology companies debate whether the soaring costs associated with cutting-edge AI development can be justified.

DeepSeek’s announcement arrives amid heightened scrutiny of the economics of large language models—tools that require extensive computing resources for training and operation. By positioning its latest model as both more capable and lower-cost, the startup is seeking to challenge the prevailing assumption that the best-performing AI necessarily demands the largest budgets and most advanced infrastructure.

The firm has not, in the information provided, detailed the specific tests used to support its comparisons, but it framed the upgrade as a meaningful step beyond its September release. Even so, the company’s emphasis on performance “by some measures” suggests the results depend on the chosen benchmarks, a common point of contention in the AI industry where models can excel in one category while lagging in others.

The development also lands at a moment when Chinese AI companies are working to demonstrate competitiveness with U.S. leaders despite differences in access to top-end chips and other constraints. A model that can deliver higher quality outputs at lower expense would be attractive not only to developers building applications on top of foundational models, but also to enterprises weighing whether to adopt AI tools at scale.

DeepSeek’s expansion into an improved offering signals intensifying competition in the generative AI market, where startups and tech giants alike are racing to deliver better accuracy, reasoning, and responsiveness, while also lowering inference costs—the price of generating each response.

Why it matters

If DeepSeek’s claims hold up under independent testing, it could put pressure on AI leaders to cut prices or accelerate model improvements, benefiting businesses and consumers looking for affordable tools. It may also strengthen China’s position in the global AI race by showing that competitive performance can be achieved with tighter cost structures. The next key test will be whether external evaluations confirm the reported gains and how quickly developers adopt the model in real-world products.