A GLM-5.3 leak says days away — we reported 5.3 was being skipped, and we are not moving

This one contradicts our own earlier reporting, so here are both sides and our read.

The claim on the card: GLM-5.3 is Z.ai's next frontier model, expected within days. Approaching Kimi K3 and GPT-5.6 Sol. 1T+ total parameters. Remaining open-weight. Built around coding, reasoning and autonomous agents. Z.ai founder Tang Jie teased an "epic" upgrade last month and this is said to be it.

What we published previously, and still think is better supported: GLM-5.5 may skip 5.3 and 5.4 entirely. That came from a JPMorgan research note relayed by Reuters on June 25, pointing at an August target for a 1T+ open-weight model with a 1M context window — the same specification the card above attaches to 5.3.

Notice that the two stories share every technical detail and disagree only about the number after the decimal point. That is the signature of one set of leaked specs being attached to two different version names by different people, not of two models.

On the evidence: GLM-5.3 as a name surfaced through community teases around Tang Jie's July 14 "are you ready?" post and his July 20 "epic upgrade" comment — neither of which named a version. Z.ai has not announced GLM-5.3, published specs, or given a date. The zai-org organisation on Hugging Face has nothing newer than GLM-5.2 from early July. So GLM-5.3 currently exists as a community expectation attached to a fast release cadence, while GLM-5.5 exists as a bank's research note relayed by a wire service.

Neither is confirmation. But a JPMorgan note via Reuters is a higher grade of source than an unattributed community expectation, and we are not downgrading our earlier call on the strength of a card.

What we would actually bet on is the part both stories agree on: Zhipu ships a 1T-class open-weight model in this window, focused on coding agents, and the version number is the least reliable thing anyone is telling you about it. Our watch entry deliberately matches the whole 5.3-through-5.9 range for that reason — whichever digit lands, it fires.

Grade: rumor, and specifically a rumor we think has the version number wrong.