ZMIME

Olala Mini 7A1B vs TML-d24-varloop-research

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

Model specifications
SpecificationOlala Mini 7A1BOVHcloudTML-d24-varloop-researchThinking Machines
TaskMultilingual language generationVariable-depth reasoning research
Published capabilitiesCompact mixture-of-experts model combining hybrid attention and Mamba layers.Researchers can vary the number of middle-layer passes with a single checkpoint.
Model size7B total / 1B active26.4B total / 3.4B active parameters
AccessDownloadable checkpoint for local inference or fine-tuningDownloadable research checkpoint
Operating limits—Native mixed-precision checkpoint occupies approximately 20 GB.
SourcesModel weights · Additional documentationModel weights · Additional documentation
InputsTextText, Images
Context window32,000 tokens—
WeightsOpen weights · Apache-2.0Open weights · Apache-2.0
ReleasedNov 19, 2026Nov 20, 2026