Flower

Flower is an open source framework for federated AI, in which models are trained across data that stays where it already lives rather than being pooled in one place. The project supplies the libraries and infrastructure to define a training strategy, connect participating nodes and run federated training and fine tuning, which makes it a route to working with data that cannot legally or practically be centralized. It is aimed at machine learning engineers and research teams in regulated or distributed settings, and deployments span industries and scales. Alongside the open framework, Flower Labs builds Flower SuperGrid, a managed service for running federations with a web interface, unlimited public and private federations, audit logs and role based access control, plus a collaborative agent and its own models. Courses on federated learning and on federated fine tuning of large language models are published through DeepLearning.AI, and the company offers forward deployed engineers for customer projects. SuperGrid pricing is published in credit tiers billed yearly: €20 per month for 8,000 monthly credits, €50 per month for 20,000 monthly credits, and €200 per month for 80,000 monthly credits, with larger allowances selectable on the same Pro plan. SuperGrid Max is contact sales, with custom credits plus advanced access control, verified apps, confidential compute, private apps on Flower Hub and SOC 2 reports. The page states that all prices include applicable VAT, and a free account comes with 3,000 sign up credits. The framework itself remains free to run on your own infrastructure. The paid tiers cover the hosted grid and its governance features rather than the library.