Alibaba just introduced Continuous-Time Distribution Matching (CDM), a distillation technique that accelerates image diffusion models down to 4 sampling steps. Unlike predecessors like DMD2 that often produce waxy, over-smoothed results, CDM preserves fine details without complex auxiliary reward models. It brings near real-time, high-fidelity generation to standard GPUs.
Stability AI just launched Stable Diffusion 3 and its Turbo version on its Developer Platform API, partnering with Fireworks AI for enterprise-grade availability. Powered by the new MMDiT architecture, the model rivals DALL-E 3 and Midjourney v6 in typography and prompt adherence. Model weights will also be available soon for member self-hosting.
Stability AI just lost the core research team behind Stable Diffusion, including lead researcher Robin Rombach and two key co-authors. The high-profile departures follow a string of executive exits, mounting compute and payroll debts, and stalled fundraising efforts, putting the open-source darling under immense survival pressure.
Midjourney just slapped an indefinite ban on all Stability AI employees after bot-like scraping brought its service down for 24 hours. The rival was allegedly trying to harvest prompt-image pairs to train its own models. Now, Midjourney is enforcing a nuclear policy: crash our servers with aggressive automation, and your entire company gets banned.
Stability AI just announced Stable Diffusion 3 in early preview, calling it its most capable text-to-image model yet. The suite ranges from 800M to 8B parameters and combines diffusion transformer architecture with flow matching to solve messy spelling and multi-subject prompts. A waitlist is now open ahead of an open release.
