Fireworks AI

Fireworks AI is a platform for companies that want to run open models, combine them across tasks, and train them on their own data. It is built for teams that need their own model system instead of renting closed frontier models.

The platform includes serverless inference, reserved throughput, and private deployment options. It supports modes like Standard, Priority, Fast, Batch, and Spot, with features such as cached input pricing, 30+ regions, and policy-aligned data residency.

It also includes training and model ensembling for building domain-specific systems. The company positions the product for enterprise AI use cases such as code assistance, conversational AI, agentic systems, search, multimodal apps, and enterprise RAG.