I'm trying to wrap my head around how the pricing for Azure AI Foundry actually works. I'm unsure if it charges me per token like some other inference services do, if there are deployment fees, or if it's a combination of both. Also, I can't seem to find specific pricing details for the input and output tokens of each model in the catalog. Am I missing something about Azure AI's pricing structure or just looking in the wrong spots?
4 Answers
You can find more details in the official pricing guide here: https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/final/en-us/microsoft-product-and-services/azure/pdf/ms-azure-ai-foundry-pricing-guide-e-book-final.pdf
I just wanted to jump in to also learn about this since we're considering using Azure AI Foundry. Would love to see more insights into the pricing model!
Just a note that Azure AI Foundry is geared towards self-hosting models, which can also affect pricing.
If you’re mainly using Azure AI Foundry to deploy OpenAI's LLMs, typically it's charged based on the number of tokens you use. However, Azure AI Foundry offers various resources that come with different pricing schemes. Most services are priced based on consumption, though some might charge by the hour for compute resources, depending on how you manage them. Services like document intelligence can charge per page or word. Just a heads up, Azure might set up several resources for you automatically which can make tracking costs tricky if you're funding it yourself, so be sure to keep an eye on the spending dashboard!
I checked the guide, but I still can’t find pricing info on models like Phi-4, Grok, or Mistral—it's mostly just focused on OpenAI models.