SenseNova Loop-9B-Preview: a 9B open model that runs its layers twice for deeper reasoning
SenseTime's open-source multimodal model, released 30 October, accepts text and images and generates text for long-context and agentic workflows; a block of 24 of its 32 decoder layers runs twice with shared weights, giving 56 layer computations from about 9 billion parameters, under Apache 2.0.
Why it matters
Scaling compute depth instead of model size: 9B of weights claim 27B-class results on vendor benchmarks (66.89 on WorkBuddy Bench against 60.53 for Qwen3.5-27B, vendor numbers not independently verified), with 262,144-token training sequences.
What you could build with it
A document-automation startup could run this 9B model on a single GPU to read long contracts and filings, draft summaries, and call tools over multi-step workflows, keeping client documents on its own servers instead of sending them to a hosted frontier API.
Does it hold up?
Usable with caveats on day one: official Transformers inference works, but the GGUF port needs a patched llama.cpp build, stock Ollama and LM Studio cannot load it yet, and the technical report is still unpublished.
Built with SenseNova Loop-9B-Preview
- SenseTime open-sources 9-billion-parameter AI model that reuses layersTechNode · Release report on the looped architecture, 262,144-token training setup, and Apache 2.0 weights plus inference and training guidance.
- SenseTime open sources SenseNova-Loop-9B-Preview, a model that runs its layers twiceTechCityAuthority · Explains the manifold-constrained hyper-connections, the 56-from-32 layer math, and the missing technical report and in-training Dspark/Dflash2 speedups.
- AtomicChat/SenseNova-Loop-9B-Preview-GGUFHugging Face · Third-party GGUF quantizations verified with a patched llama.cpp build, noting stock llama.cpp, LM Studio and Ollama do not yet support the architecture.
Learn more
First spotted on hf: source.
More AI models
VoiceStudio: open-source, fully-local ElevenLabs alternative for voice cloning, dubbing, and audiobooksdebpalash/VoiceStudio is a desktop app that runs voice cloning, voice design, video dubbing, dictation, transcription…models · JEV 0.75Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72NVIDIA Cosmos 3.5 world models: now openly availableNVIDIA's Cosmos 3.5 family of open world foundation models for robots, autonomous vehicles and vision agents is now…models · JEV 0.71
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.