ZMIME

Hy4 Base vs Olala Mini 7A1B

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

Model specifications
SpecificationHy4 BaseTencentOlala Mini 7A1BOVHcloud
TaskMultimodal pretrainingMultilingual language generation
Published capabilitiesPretrained checkpoint for continued training and adaptation. Tencent’s flagship Hy4 uses a 78-layer language backbone with a separate vision encoder.Compact mixture-of-experts model combining hybrid attention and Mamba layers.
Model size770B total / 49B active per token in the language backbone7B total / 1B active
AccessDownloadable under the Tencent Hy LicenseDownloadable checkpoint for local inference or fine-tuning
SourcesModel weightsModel weights · Additional documentation
InputsText, Images, VideoText
Context window—32,000 tokens
WeightsOpen weights · Tencent Hy LicenseOpen weights · Apache-2.0
ReleasedNov 20, 2026Nov 19, 2026