NodeValid

Published video-model weights, licences and hardware notes

HunyuanVideo says over 13 billion

The size here is written as over 13 billion parameters. It is the vendor's own phrasing, and it is a bound rather than a count, which puts this row in a different class from the releases that publish a number. As of 2026-09-12.

HunyuanVideo in the size column, as publishedA bound tells a reader which order of magnitude the download sits in and refuses the one comparison the size column exists to support.Fixed by the published countLeft to the runThe countOver 13 billion parametersWhat it belongs toStated in the repositoryLicence on that releasetencent-hunyuan-communityMemory for a run60GB peak at 720px1280px, 129 framesOutput sizeNo output size recordedA bound, not a count
Fig. 1 An open-ended figure sets a floor for the file and leaves the comparison undone.
HunyuanVideo in the size column, with the columns that qualify it. Read from github.com/Tencent-Hunyuan on 2026-09-12.
ColumnWhat the pages state
Size, in the vendor's own wordsOver 13 billion parameters
The detail published beside itStated in the repository
Licence, from the same pagestencent-hunyuan-community
Hardware, from the same pages60GB peak at 720px1280px, 129 frames

Inclusion rule. The value this page is about, the detail the vendor attached to it, and the two columns that change how it should be read. A column the vendor left empty keeps its row and says so. Order. This column first, then its detail, then the two columns that qualify it.

1Why the hedge is not carelessness

Research releases often count differently: text encoders in or out, the autoencoder separately, embeddings sometimes ignored. Writing over thirteen billion avoids committing to a convention that the next paper might not share.

It is a reasonable defensive choice for an author and an inconvenient one for a register, which needs the figure precisely so that two rows can be put beside each other.

2What the bound still buys

Enough to plan storage and to place the release. A model above thirteen billion parameters is not going to be the one that runs on a laptop, and the memory figures on the same pages confirm it at 60GB and 45GB.

So the size column and the hardware column agree here, which is not always the case. Agreement between two columns is worth noticing precisely because it is not guaranteed.

3The comparison this row cannot join

One release in the register publishes its count to the individual parameter. Putting a bound beside an exact figure produces a comparison that looks valid and is not, because one side has an unknown upper end.

The register keeps the phrasing as written. Rounding a bound into a number to make the column tidy would manufacture a figure nobody published.

4Sources

The count comes from github.com/Tencent-Hunyuan, read 2026-09-12, exactly as the vendor wrote it. Everything else this release publishes is on its family entry; why the count sits in its own column is on size. Nearby: Size in the model name, Counted in two parts.