ZMIME

Aion 1.0 Instruct 2.3B vs Inkling 2 Mini

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

Model specifications
SpecificationAion 1.0 Instruct 2.3BMicrosoftInkling 2 MiniThinking Machines
TaskOn-device language generationMultimodal reasoning and coding
Published capabilitiesWindows-oriented instruction model for local text generation, including document and email tasks.Reuses its middle layers three times per token. Thinking Machines reports 48.2% on DeepSWE v1.1.
Model size2.3B parameters26.4B total / 3.4B active parameters
AccessDownloadable checkpointDownloadable weights and Tinker inference and customization
Operating limits32K-token context windowNative mixed-precision checkpoint occupies approximately 20 GB.
SourcesModel weightsModel weights · Additional documentation
InputsTextText, Images
WeightsOpen weights · MITOpen weights · Apache-2.0
ReleasedNov 18, 2026Nov 20, 2026