ZMIME

Hy4 flash Base vs Hy4 Base

Compare documented capabilities and access. A dash means a comparable figure is unavailable.

Model specifications
SpecificationHy4 flash BaseTencentHy4 BaseTencent
TaskMultimodal pretrainingMultimodal pretraining
Published capabilitiesPretrained checkpoint for continued training and adaptation. The Flash architecture uses a sparse Engram memory module.Pretrained checkpoint for continued training and adaptation. Tencent’s flagship Hy4 uses a 78-layer language backbone with a separate vision encoder.
Model sizeApproximately 220B total / 12B active per token770B total / 49B active per token in the language backbone
AccessDownloadable under the Tencent Hy LicenseDownloadable under the Tencent Hy License
SourcesModel weightsModel weights
InputsText, ImagesText, Images, Video
WeightsOpen weights · Tencent Hy LicenseOpen weights · Tencent Hy License
ReleasedNov 18, 2026Nov 20, 2026