Union

Union is an AI runtime for teams that need to run durable AI and machine learning workflows inside their own cloud infrastructure. It uses Python-based workflows and the Flyte open-source platform, supports runtime logic, conditions, retries, caching, and live remote debugging, and tracks usage costs. Interrupted workloads can resume instead of starting over. It also handles real-time inference and compute optimization while keeping sensitive data in the customer infrastructure. The Team plan includes one cluster, 1,000 concurrent actions, fanout up to 25,000 actions per run, 30-day retention, OIDC SSO, role-based access control, managed secrets, and dedicated support. Enterprise raises or customizes those limits, adds at least three clusters, one-year retention, finer access controls, and custom support. Pricing combines a monthly usage credit with metered actions and allocated CPU, memory, or GPU resources, so buyers should estimate both workflow volume and resource mix.