SOURCE-REVIEWED MODEL PROFILE
SenseNova-Loop-9B
A text-and-image model that reuses a middle block of decoder layers during inference.

Metadata reviewed 2026-11-19 · Original creator: SenseNova
PARAMETERSApproximately 9B
CONTEXT512K tokens per model card
LICENSEApache-2.0
MODALITYImage and text to text
What it does
- Looped decoder
- Vision and text inputs
- AWQ and GGUF options
Local fit
See upstream runtime and memory requirements. Hugging Face access step applies to the main checkpoint. Runtime support and memory depend on the variant.
Publisher specification. Check the exact variant and runtime.
Evidence and limits
Source reviewed. No independent hands-on test.
Source reviewed
License and source
Hugging Face access step applies to the main checkpoint. Runtime support and memory depend on the variant.
Read the original model card ↗
Reviewed source revision: 4602ecf7e8b023ec081572f431560d8a698bf0cc