Introducing Fireworks AI on Microsoft Foundry: Bringing high performance, low latency open model inference to Azure

Microsoft is announcing the public preview of Fireworks AI on Microsoft Foundry, integrating Fireworks AI's high-performance open model inference engine into Azure's enterprise AI platform. Fireworks AI processes over 13 trillion tokens daily at ~180K requests per second and generates over 1,000 tokens/second on large models. Through this integration, developers can access models including DeepSeek V3.2, OpenAI gpt-oss-120b, Kimi K2.5, and the newly added MiniMax M2.5 via a single Azure endpoint. Key capabilities include serverless pay-per-token inference, provisioned throughput units (PTUs), bring-your-own-weights (BYOW) support for custom fine-tuned models, and enterprise governance through Foundry's unified control plane.