SenseNova-Loop-9B-Preview TRIAL

Public weights and code make looped multimodal execution testable, while serving support is still landing.

Next stepWithin one day, compare fixed and looped compute on public examples for accuracy, latency and memory.

Judgment reviewed 2026-10-31 · Basis: SenseNova opens a looped multimodal model for long-context agents

RSS

Availability & adoption

availability: Ungated BF16 weights and public code under Apache-2.0

Timeline

2026-10-30New

SenseNova opens a looped multimodal model for long-context agents

Official · ReferenceSenseNova-Loop-9B-Preview model card

Building agents →Local deployment →Choosing models →