Skip to main content
← Back to news

SenseNova-Loop-9B: the looped architecture that turns 32 layers into 56 executions with 9B parameters

Model releases
SenseNova's multimodal model reuses the weights of its 24 middle layers twice (looped block with two-stream mHC hyper-connections): context up to 512K tokens and, with 4-bit weights and an 8-bit KV cache, roughly 15 GiB of VRAM for a 256K context. GAIA 69.90, Terminal-Bench 2.1 59.55; on AIDABench it beats Qwen3.8-27B (65.31 vs 65.01).

Get the next local-AI briefing

Once a week, the top selection from iaLocale: articles, news and resources on local AI, and more.

After signing up you receive a confirmation email (double opt-in). Without confirmation we send nothing. No spam, we promise.