sensenova/SenseNova-Loop-9B-Preview
AI model · https://huggingface.co/sensenova/SenseNova-Loop-9B-Preview
First seen by Oriel on .
AI summary
SenseNova-Loop-9B-Preview is a 9B-scale multimodal decoder using shared looped layers and two-stream mHC for long-context and agentic workflows.
- The model accepts text and images and generates text for document understanding, coding, and external-tool interaction.
- A shared block of 24 decoder layers runs twice, producing 56 decoder-layer executions from 32 distinct layer weight sets.
- Hybrid attention repeats three linear-attention layers followed by one full-attention layer, while mHC routes information across two residual streams.
- Training recipes support sequences up to 262,144 tokens and combine text, images, code, and supervised tool trajectories.
- MTP is available for inference acceleration; Dspark and Dflash2 remain under development, and deployment requires compatible custom model code and execution parameters.
#multimodal models #long context #agentic workflows #transformers
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 286, latest observation ; +183 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 3,870, latest observation ; +3,304 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | sensenova/SenseNova-Loop-9B-Preview | https://huggingface.co/sensenova/SenseNova-Loop-9B-Preview Read discussion |