Reported GLM-5.5 pipeline claims 1 trillion crawled web pages for Z.ai
A November 4th post attributes six internal data engines to Jie Tang's GLM-5.5 work, with no official Z.ai announcement confirming the figures.
By Ryan Merket · Published
Primary source: x.com
Why it matters
The post asserts that Z.ai is building pretraining data in-house across web, documents, books, code, video and audio at extreme scale. If Z.ai confirms even part of that operation, data supply becomes a clearer competitive axis for its next flagship. Until then, the key fact is attribution: the sizes previously circulated as internal checkpoints, and Z.ai's shipped flagship remains GLM-5.3.
Jie Tang, a co-founder of Z.ai and a Tsinghua University professor, was reported on November 4th to have described the pretraining pipeline behind an upcoming GLM-5.5 during a lecture at Tsinghua University. The account came from bebu (@bebuu_ai) in a post on X, which attributed six internal data engines to the project and listed model sizes of 32B, 122B and 256B.…
When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.