ZMIME

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

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

Model specifications
SpecificationSage S2.1 FlashShanghai AI LaboratorySenseNova-Loop-9BSenseTime
TaskMultimodal scientific reasoningVision-language reasoning
Published capabilitiesAnalyzes scientific inputs including molecular graphs, crystal structures and protein sequences, using tools to verify intermediate results.Reuses middle layers to perform 56 layer executions with 32 physical decoder layers.
Model size321B parameters9B parameters
AccessPublic checkpoint with vLLM and SGLang supportPublic weights including quantized variants for local inference
SourcesModel weightsModel weights
InputsText, ImagesText, Images
Context window—524,288 tokens
Input price per million tokens——
Output price per million tokens——
WeightsOpen weightsOpen weights
ReleasedNov 17, 2026Nov 17, 2026