Partly True
A claim that the interval between new AI model releases collapsed from 210 days to 47 days within a single year โ a figure cited in Korean coverage of the intensifying race to ship large language models โ has been put through a source-by-source verification that traced it to a Dong-A Ilbo article and weighed 12 primary sources before landing on a split finding.
The claim originates in an article published on donga.com, the online platform of the Dong-A Ilbo. During verification, the original URL could not be surfaced, so reviewers confirmed the report's contents through re-publications carried on Naver and Nate. The syndicated copies matched on the central figures: an average gap of 210 days between successive new model disclosures in the earlier period, shrinking to 47 days a year later.
The two numbers anchor the acceleration argument. A release cadence measured in months shortening to one measured in weeks would mark a sharp compression of the development cycle, and the claim frames that shift as evidence that the pace of AI model rollouts is quickening year on year. It is the kind of stat that has become a staple of coverage as Korean and global developers push out LLM updates at an ever-faster clip.
Reviewers examined 12 primary sources connected to the claim, from the original print article to the Naver and Nate re-publications used to confirm its contents. The exercise did not resolve every element cleanly: portions of the acceleration narrative held up against the record, while others could not be fully substantiated, leaving the claim supported in part rather than in whole. Measured against the Dong-A Ilbo source trail and the 12 primary sources, the verification rated the 210-to-47-day acceleration claim as Partly True.