Plans

Featherless provides serverless access to AI models and agent runtimes, allowing you to power AI applications without managing your own inference infrastructure.

Featherless offers two primary types of plans:

Feather Chat plans are subscription and concurrent-unit based, providing unlimited monthly requests with a fixed number of concurrent units. Subscription tiers vary by supported model sizes, context lengths, and the number of concurrent units included. These plans range from smaller consumer tiers designed for interactive chat, assistants, and role-playing to larger plans built for coding and agentic workloads.

Feather Developer plans are subscription and credit based. Instead of unlimited requests tied to concurrent units, usage is deducted from a monthly credit balance based on the models and inference consumed. Developer plans are designed for API-driven applications and workloads where usage may vary over time.

Chat Plans

*smaller models allow for higher concurrency than larger models. See more below.

Developer Plans

Developer plans are scalable, allowing users to purchase larger amounts of inference to for coding or to power production applications - whether agent fleets or other AI applications.

Dedicated GPU

Dedicated GPU deployments are available for workloads that need more capacity, custom infrastructure, or greater control than our standard plans provide. They are a good fit for larger models, specialized configurations, consistently high-throughput workloads, or applications whose growth has reached the point where token-based billing is no longer the most economical option.

Dedicated deployments can be tailored around your model, performance, and scaling requirements, giving you reserved capacity and predictable infrastructure costs.

*For more info on how the concurrent unit limits work visit: