GPT-4o mini: now available on Azure AI

[justinroyal]

GPT-4o mini by OpenAI is now available on Azure AI. This new model is touted to be smarter and more cost-effective than its predecessor, GPT-3.5 Turbo, boasting an 82% score on the MMLU compared to 70%, and offering a 60% cost reduction. It features a 128K context window and improved multilingual capabilities, enhancing quality across various languages.

GPT-4o mini supports text processing on Azure AI with image, audio, and video capabilities to be added later. It is particularly beneficial for streaming scenarios like assistants, code interpreters, and retrieval services due to its speed and efficiency. The model's integration with GitHub Copilot has demonstrated remarkable speed, providing code completion suggestions almost instantaneously.

Azure AI has also introduced updates to the Azure OpenAI Service, focusing on safety, data residency, and pay-as-you-go availability. Safety features such as prompt shields and protected material detection are now enabled by default. The service now offers data residency in all 27 regions, including the newly launched region in Spain, ensuring compliance with customers' unique requirements.

The global pay-as-you-go deployment option for GPT-4o mini is now generally available, offering competitive pricing and high throughput limits. Customers can upgrade to newer models without changing regions, and the service promises 99.99% availability with industry-leading speed.

Finally, Azure AI is investing in efficiencies for AI workloads, introducing fine-tuning for GPT-4o mini and reduced hosting charges. 

Check out this blog to learn more: OpenAI’s fastest model, GPT-4o mini is now available on Azure AI | Microsoft Azure Blog

Are you already Azure AI in your app development? Comment below to let us know what additional resources would be helpful on your AI journey!

[#2]

I checked WestUS, EastUS and SouthCentralUS in my org's tenant. I do not see gpt-4o-mini as an option to deploy. How do I "turn this on"?

[#3]

Disregard. I found it by adding the following to my Chat Playground URL: `https://oai.azure.com/portal/<resource-id>/early-access-playground`

[#4]

@seadude but not available for deployment yet? only preview in playground?

[#5]

@dimahavry GPT-4o Mini is only available in early access playground, Not available for the deployment.

[#6]

Is this still this still unavailable for deployment? Any idea when it will be?

I keep looking for updates and articles like this one from Microsoft give every indication that it is but it's not. https://azure.microsoft.com/en-us/blog/openais-fastest-model-gpt-4o-mini-is-now-available-on-azure-a...

"GPT-4o mini is now available using our global pay-as-you-go deployment at 15 cents per million input tokens and 60 cents per million output tokens, which is significantly cheaper than previous frontier models."

I don't necessarily mind if it's a bit late to the party but just don't waster our time telling us things that aren't true.

[#7]

Like everyone else I am wondering when it will be available for deployment and the costs.

[#8]

Hello @dfcarter @JamesP250 - API and Fine-tuning for GPT-4o mini is available now. You can refer to this more recent blog for additional details:

OpenAI’s GPT-4o mini Now Available in API with Vision Capabilities on Azure AI (microsoft.com)

[#9]

Good on you @justinroyal

[#10]

@justinroyal great! Thanks for the update 

[#11]

I just checked my zone, West Europe. And it seems that GPT-4o-mini has yet to become available. I do wonder when this will happen, because GPT-3.5-turbo is being deprecated in 2 weeks