SenseTime open-sources SenseNova U1, a series of native multimodal models that integrate image and text understanding, reasoning, and generation using the NEO-unify architecture.
Models post state-of-the-art open-source benchmark results with interleaved image-text output support.
1/ I have been spending time with SenseNova U1, a native multimodal model series released by @SenseTime_Al.
It is built on an architecture called NEO-unify that processes images and text together in one single system. It is a big change from the usual way of handing tasks off between separate components.
Look at this thread 🧵:
2/ I tested the interleaved generation by starting with a monochrome sketch of Darth Vader. It builds out mechanical textures and lighting step by step while keeping the same visual style throughout.
Watching it refine the image without losing the original structure is pretty impressive..
Chubby♨️@kimmonismus1/ I have been spending time with SenseNova U1, a native multimodal model series released by @SenseTime_Al. It is built on an architecture called NEO-unify that processes images and text together in one single system. It is a big change from the usual way of handing tasks off between separate components. Look at this thread 🧵:
3/ It also handles high-density info rendering very well. I generated this awareness infographic to see how it manages complex layouts.
It kept the text clean and the icons structured, which is a common pain point for open-source models.
Chubby♨️@kimmonismus2/ I tested the interleaved generation by starting with a monochrome sketch of Darth Vader. It builds out mechanical textures and lighting step by step while keeping the same visual style throughout. Watching it refine the image without losing the original structure is pretty impressive..
6/ The full Technical Report is now out, their most detailed model disclosure yet. SenseNova-U1-A3B-MoT (38B-A3B MoE) weights are now open-sourced.
You can check out the report or try the tools at the links below.
Try it here: https://unify.light-ai.top/login?next=%2Fhome
Technical Report: https://arxiv.org/abs/2605.12500
GitHub: https://github.com/OpenSenseNova/SenseNova-U1
Hugging Face: https://huggingface.co/collections/sensenova/sensenova-u1
Chubby♨️@kimmonismus5/ They are open-sourcing the Lite series in two sizes, an 8B dense model and an A3B mixture-of-experts version. The product has some exciting new updates. An 8-step distilled LoRA now open-sourced: inference cut from 23s to 2s on H100 (100 NFE -> 8 NFE). ComfyUI is now supported, with ready-to-run workflows for text-to-image, image editing, and interleaved generation.