SenseTime Releases SenseNova Loop-9B Model: Deployment VRAM Requirement as Low as Under 10 GB
IT Home news on November 17: SenseTime officially released the SenseNova Loop-9B model this evening (the 17th).
The model continues the Preview version's exploration of parameter efficiency: by reusing model layer parameters through the Loop structure and combining it with manifold-constrained hyper-connections (mHC), it extends computational depth at a scale of roughly 9B parameters, targeting complex reasoning, code, tool calling, and multimodal tasks.
Thanks to the Loop's efficient recurrent structure, the official 4bit-quantized weights are only about 4.9GB, and the deployment VRAM requirement is as low as under 10 GB, significantly broadening the range of deployable hardware and making it possible for low-VRAM chips to achieve frontier 27B-level model performance.
SenseNova Loop-9B has a total parameter count of approximately 9B. By executing the shared intermediate layers twice, the model expands 32 physical decoder layers into 56 layer computations; it supports text and image input and text output, with a context length of up to 512K tokens.
According to the official statement, in Agent and data analysis benchmarks, the model demonstrates competitiveness comparable to Qwen3.8-27B: it scores 41.36 on NL2Repo, close to the latter's 44.60; and 65.31 on AIDABench, higher than the latter's 65.01.
- Shared recurrent module: layers 4–27 form a weight-sharing recurrent module and are executed twice, enabling the model to complete 56 decoder layer computations in a single forward pass.
- Dual-stream mHC: through learnable mappings, information routing and fusion are performed on the two residual streams in the recurrent module.
- Hybrid attention: the decoder uses groups of four layers, adopting a configuration that combines three layers of linear attention with one layer of full attention.
- Long-context multimodality: the training scheme covers text, images, code, and tool-calling trajectories, with context length extended up to 512K tokens.
- Friendly to low-resource private deployment: with 4bit weight quantization and 8bit KV cache, the VRAM requirement for 256K long-context inference is approximately 15 GiB, helping lower the barrier to local and private deployment.
- Official quantized version available on Day 0: an official AWQ 4bit quantized version is provided on the first day of release, making it convenient for users to deploy quickly.
SenseNova Loop-9B has released its model weights and related code for developers and researchers to download, deploy, and explore.

Around tasks such as general Agents, code, and data analysis, benchmark results show that the official version of SenseNova Loop-9B has significantly improved performance on multiple leaderboards compared with the Preview version; and at a scale of approximately 9B parameters, it demonstrates performance comparable to Qwen3.8-27B.
IT Home attaches the Hugging Face repository: https://huggingface.co/collections/sensenova/sensenova-loop
Advertising statement: External redirect links contained in this article (including but not limited to hyperlinks, QR codes, passwords, and other forms) are used to convey more information and save selection time; the results are for reference only. All articles on IT Home that contain external links include this statement.
Related Articles
- SenseTime's AI office agent "Xiaohuanxiong" mobile app launches, offering Apple iOS / Android versions
- SenseTime SenseNova 6.8 Flash Lite preview version launches: focusing on lightweight agility, aiming at "delegated intelligence"
- Native support for 4K image generation, SenseTime open-sources multimodal model SenseNova U1.5-Lite-Preview preview version
- SenseTime releases SenseNova U1 Pro image creation model: up to 8K native resolution output, claims to say goodbye to the "AI feel"
- Reportedly SenseTime is developing a new image model: codenamed U1 Pro, internally said to benchmark against OpenAI
- SenseTime and Hong Kong Science and Technology Parks launch cooperation to jointly build Hong Kong's largest domestic intelligent computing center