Model releaseinfo
SenseNova-Loop-9B-Preview multimodal model released
Original title: sensenova/SenseNova-Loop-9B-Preview — new model trending #30 on Hugging Face
AI summary · x-ai/grok-4.7Summaries by Vani AI
New 9B-parameter looped model handles text, images and up to 262K tokens for agentic workflows.
SenseNova-Loop-9B-Preview is a new multimodal decoder released by OpenSenseNova. It uses a shared-weight looped block and two-stream mHC to extend computation depth without parameter duplication. The model supports long contexts and external tool interactions after training on multimodal trajectories.
Key facts
- 24 decoder layers executed twice for 56 total passes with shared weights
- Maximum training context length 262144 tokens
- Hybrid attention interleaves linear and full attention every four layers
- ~9B total parameters with ~14.6B inference cost
- Trained on text, images, code and tool-use examples
- Organizations
- OpenSenseNova
- AI models
- SenseNova-Loop-9B-Preview
From the original article
6 min read at huggingface.coimage-text-to-text · 927 downloads · 150 likes · license apache-2. 0
By sensenova · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.