SenseTime open-sources SenseNova-Loop-9B-Preview, a looped multimodal model that runs a shared 24-layer block twice (Apache 2.0)
Confirmed
The takeaway
SenseTime released SenseNova-Loop-9B-Preview on Oct 30, 2026: an Apache-2.0, ~9B-parameter text-and-image model whose middle block of 24 decoder layers is run twice, adding depth of computation without extra weights.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 19 November 2026
Your AI and this story
- GPT-6.2 Astra152 days after its cutoff
- Claude Opus 5.5122 days after its cutoff
- Gemini 4 Hydrogen213 days after its cutoff
- Grok 4.7152 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 122 days before it.
Key facts
- ~9B parameters; shared 24-layer block executed twice
- Input text and images, output text; aimed at document understanding, coding and tool use
- Weights, inference instructions and training guidance released; licence Apache 2.0
- Nov 18 (TechNode): SenseTime released Loop-9B as an open-weight model; it reuses layers in a loop for deeper computation
Sources
4 sources from 3 sites. Numbers match the chips in the text.
4 sources: 1 primary, 3 press
Primary
- GitHub: OpenSenseNova/SenseNova-Loopgithub.com, code
Press
- TechNode: SenseTime releases open-weight Loop-9Btechnode.com, press
- TechNode (Nov 2)technode.com, press
- Pandaily (Nov 1)pandaily.com, press
Changes
- Open-weight release
- Filed