ZMIME

SenseNova-Loop-9B vs Sage S2.1 Flash

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

Model specifications
SpecificationSenseNova-Loop-9BSenseTimeSage S2.1 FlashShanghai AI Laboratory
TaskVision-language reasoningMultimodal scientific reasoning
Published capabilitiesReuses middle layers to perform 56 layer executions with 32 physical decoder layers.Analyzes scientific inputs including molecular graphs, crystal structures and protein sequences, using tools to verify intermediate results.
Model size9B parameters321B parameters
AccessPublic weights including quantized variants for local inferencePublic checkpoint with vLLM and SGLang support
SourcesModel weightsModel weights
InputsText, ImagesText, Images
Context window524,288 tokens—
Input price per million tokens——
Output price per million tokens——
WeightsOpen weightsOpen weights
ReleasedNov 17, 2026Nov 17, 2026