US research institution: DeepSeek's latest model costs far less than rivals

On August 3, Lianhe Zaobao cited a U.S. research institution's report finding that DeepSeek's latest model carries costs far lower than other models, putting DeepSeek's price advantage back in the industry spotlight.

The core of the report is the cost comparison itself: among comparable models, DeepSeek's latest model is significantly cheaper, with no specific figures disclosed in the coverage.

For the AI industry, model cost directly shapes inference pricing and the economics of large-scale deployment, and cheaper models tend to reach high-frequency business scenarios faster.

For developers and enterprises, lower model costs ease inference budget pressure and make more long-running, high-concurrency AI applications economically viable.

The finding arrives as global model vendors compete aggressively on price and performance, making cost data a key measure of a model's practical value.

Worth watching next are the specifics and methodology of the research institution's report, and whether DeepSeek extends this cost edge to its next-generation models.