cd /news/artificial-intelligence/sensetime-releases-open-weight-loop-… · home › topics › artificial-intelligence › article
[ARTICLE · art-171633] src=technode.com ↗ pub= topic=artificial-intelligence verified=true sentiment=↑ positive

SenseTime releases open-weight Loop-9B model that reuses layers for deeper computation

SenseTime released the weights and code for SenseNova Loop-9B on Tuesday, a roughly nine-billion-parameter multimodal model that runs its shared middle layers twice to perform 56 layers' worth of computation per forward pass from 32 decoder layers. SenseTime says the model, which accepts text and images, supports reasoning, coding, data analysis and tool use with a 512,000-token context window, is competitive with Qwen3.8-27B on selected agent and data-analysis benchmarks. An official four-bit quantized version cuts weights to about 4.9 GB, and running it with a 256,000-token context and eight-bit KV cache requires about 15 GiB of memory, according to the model documentation.

key takeaway

Loop-9B executes its shared middle layers twice to get 56 layers' worth of computation from 32 decoder layers, and its official four-bit quantized weights take about 4.9 GB, with a 256,000-token context and eight-bit KV cache needing about 15 GiB of memory.

read1 min views16 publishedNov 18, 2026
SenseTime releases open-weight Loop-9B model that reuses layers for deeper computation
Image: Technode (auto-discovered)

SenseTime released the weights and code for SenseNova Loop-9B on Tuesday, a multimodal model that increases computational depth by running some of its layers more than once. The roughly nine-billion-parameter model accepts text and images, produces text, and supports reasoning, coding, data analysis and tool use. Its context window extends to 512,000 tokens.

The model has 32 decoder layers, but executes its shared middle layers twice to perform 56 layers’ worth of computation in each forward pass. SenseTime says it is competitive with Qwen3.8-27B on selected agent and data-analysis benchmarks. An official four-bit quantized version reduces the weights to about 4.9 GB. Running it with a 256,000-token context and an eight-bit KV cache requires about 15 GiB of memory, according to the model documentation. [SenseTime]

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @sensetime 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/sensetime-releases-o…] indexed:0 read:1min 2026-11-18 · —