SenseTime Launches Native 8K Image Generation Model
量子位 · wechat · 2026-07-18
At WAIC 2026, SenseTime launched a new model, 日日新 SenseNova U1 Pro. The article describes it as a "delivery system" for complex multimodal tasks—capable of understanding, planning, organizing information, generating multimodally, and self-correcting to deliver a final product, rather than just generating images.
Key demonstrated capabilities include native 8K ultra-long image output, continuous interleaved text-and-image creation, and the ability to deliver finished assets for infographics, urban planning, film storyboards, academic posters, and commercial design. The author tested it with a 24 solar terms scroll, a lab-style poster, a glazed landscape painting, and a movie poster, noting solid performance in long-image details, layout stability, and complex style execution.
Technically, SenseTime reportedly uses 32×32 large patches to manage the visual token and VRAM pressure of 8K, combined with adaptive NoiseControl, patch overlapping, spatial sampling, and specific loss designs to preserve small text, textures, and structures. The article concludes that multimodal generation is shifting from "single-point image output" to "system-level content delivery," with U1 Pro representing an early form of this trend.
Related event: SenseTime Launches SenseNova U1 Pro Native 8K Multimodal Model(6 posts)→
More from Companies & People
- SF 'Code Engineering' event picks up where 1,000-person harness engineering meetup left off — LangChain · 2026-10-27
- Anthropic reveals it bans customers worth 5-7% of revenue every 9 months — Polymarket · 2026-10-27
- Anthropic's IPO Filing Goes Public, Revealing Strong Revenue and Massive Spending — KateClarkTweets · 2026-10-27
- OpenAI Foundation funds UNC tumor antigen project, hiring senior experimental scientist — iskander · 2026-10-27
- Anthropic IPO timing fuels claims anti-open-weights panic is self-serving — mark_k · 2026-10-27
- Anthropic's S-1 reveals over half of $100K+ customers now use both Claude Platform and first-party apps — TechCrunch AI · 2026-10-27