Managed APIs let developers run open-weight AI models without operating GPUs, with token-based billing and compatibility with common application frameworks.

Managed inference service runs FriendliAI's optimized serving stack on the customer's own cloud GPUs while FriendliAI handles deployment and operations.

Inference monetization platform fills idle GPU capacity with paid model-serving workloads while immediately yielding hardware when the operator's primary workloads need it.

Managed dedicated GPU endpoints serve open-source or custom models with autoscaling, deployment controls, monitoring, and capacity reserved for the customer's workloads.

Self-managed container deploys generative AI inference on customer infrastructure, giving teams control over GPU resources, networking, configuration, and scaling.

Command-line tool connects coding agents to Friendli Model APIs, supports model discovery and switching, and restores saved configurations when disabled.