ZMIME

Inkling 2 Mini vs Aion 1.0 Instruct 2.3B

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

Model specifications
SpecificationInkling 2 MiniThinking MachinesAion 1.0 Instruct 2.3BMicrosoft
TaskMultimodal reasoning and codingOn-device language generation
Published capabilitiesReuses its middle layers three times per token. Thinking Machines reports 48.2% on DeepSWE v1.1.Windows-oriented instruction model for local text generation, including document and email tasks.
Model size26.4B total / 3.4B active parameters2.3B parameters
AccessDownloadable weights and Tinker inference and customizationDownloadable checkpoint
Operating limitsNative mixed-precision checkpoint occupies approximately 20 GB.32K-token context window
SourcesModel weights · Additional documentationModel weights
InputsText, ImagesText
WeightsOpen weights · Apache-2.0Open weights · MIT
ReleasedNov 20, 2026Nov 18, 2026