Microsoft has launched a public preview of Fireworks AI on its Microsoft Foundry platform, enabling developers to run open models efficiently at enterprise scale. This integration aims to streamline the AI lifecycle by providing a unified environment for model evaluation, deployment, and governance. Fireworks AI, known for its high-performance inference capabilities, now supports popular open models such as DeepSeek V3.2, OpenAI gpt-oss-120b, Kimi K2.5, and MiniMax M2.5. The platform allows developers to access these models through a single Azure endpoint, simplifying the process of deploying and managing open models. According to Microsoft, Fireworks AI processes over 13T tokens daily and handles more than 180,000 requests per second, with benchmarked performance on Artificial Analysis. Developers can choose between serverless, pay-per-token inference or provisioned throughput units (PTUs) for predictable performance. The integration also supports bring-your-own-weights (BYOW), allowing users to upload and register quantized or fine-tuned models without altering the serving stack. Microsoft Foundry provides an end-to-end workspace for agent development, evaluation, and deployment, with unified governance and observability. *Source: [azureai](https://azure.microsoft.com/en-us/blog/introducing-fireworks-ai-on-microsoft-foundry-bringing-high-performance-low-latency-open-model-inference-to-azure/)*