Bagel introduced Paris today, a diffusion model that the team trained in a decentralized way and released with open weights. Bagel calls Paris the first model of its kind. The project targets creators and developers who want more freedom to build and ship.<\/p>\n\n\n\n
Paris is a generative model built on diffusion. People often use diffusion models to create images<\/a> and other media. Bagel released the model with open weights, so teams can run it on their own hardware and explore how it works.<\/p>\n\n\n\n Bagel says it trained Paris in a decentralized way. That approach can spread work across many machines and reduce control by a single host. It may lower costs, improve speed, and invite wider input from builders around the world.<\/p>\n\n\n\n Bagel allows both research and commercial use. Startups, labs, and artists can test Paris in products, prototypes, or studies without waiting for extra approvals.<\/p>\n\n\n\n The team chose the name to honor Paris, a city that often gives cover to makers who create without permission. The brand nods to open culture and artistic freedom.<\/p>\n\n\n\n Bagel shared the news on X under @bageldotcom. Visit bagel.com<\/a> or the post for access, setup details, and updates. We will update this story as Bagel shares more information.<\/p>\n","protected":false},"excerpt":{"rendered":"Why decentralized training matters<\/h2>\n\n\n\n
Who can use it<\/h2>\n\n\n\n
Why the name \u201cParis\u201d<\/h2>\n\n\n\n
Availability and next steps<\/h2>\n\n\n\n