SenseTime open-sources SenseNova U1.5: 8B unified model beating GPT-Image-2 on visual reasoning
liuziwei7 · x · 2026-09-17
SenseTime released the SenseNova U1.5 technical report: an open-source 8B-MoT natively unified model that goes beyond 'seeing + drawing' into genuine capability interaction. Key points: Native MoT connects understanding and generation via shared attention; Pixel Shuffle + 3×3 spatial convolutions enable native 4K generation with stronger spatial coherence; aesthetics, OCR, infographics and editing experts are consolidated via Multi-Expert On-Policy Distillation; WISE improves 0.70→0.81 with CoT; RealUnify-GEU rises 47.5→56.3; and it scores 68.2% on VBVR-Pro-Bench visual reasoning, ahead of Nano-Banana-Pro (56.4%) and GPT-Image-2 (50.7%). Full SFT, RL and distillation training recipes are open-sourced.
Related event: SenseTime Open-Sources SenseNova U1.5, an 8B Native Unified Model(2 posts)→
More from Models
- Epoch AI: open-weight GLM 5.3 exploited the patch gap to hack 6 Firefox releases at ~$1,000 each — rohanpaul_ai · 2026-11-02
- SemiAnalysis: safety classifiers sandbag scores of Opus 5.5 and GPT-6 Astra — morqon · 2026-11-02
- Liquid AI's 1.7B LFM3 becomes first local model to nail live fixtures lookup on a MacBook — NielsRogge · 2026-11-02
- Claude upgrade-page traffic rebounded to 17.6M in September, near May's 18.6M peak — firstadopter · 2026-11-02
- Open-source GLiNER-PII scrubs 13 identifiers from a discharge note in 550ms on GPU — MaziyarPanahi · 2026-11-02
- OpenAI Dots user says bloat and broken infra drove him to try newcomer Muse — Koala_Confused · 2026-11-02