We are besides announcing information features by default for GPT-4o mini, expanded information residency and work availability, positive show upgrades to Microsoft Azure OpenAI Service.

GPT-4o mini allows customers to present stunning applications astatine a little outgo with blazing speed. GPT-4o mini is importantly smarter than GPT-3.5 Turbo—scoring 82% connected Measuring Massive Multitask Language Understanding (MMLU) compared to 70%—and is much than 60% cheaper.¹ The exemplary delivers an expanded 128K discourse model and integrates the improved multilingual capabilities of GPT-4o, bringing greater prime to languages from astir the world.

GPT-4o mini, announced by OpenAI today, is disposable simultaneously connected Azure AI, supporting substance processing capabilities with fantabulous velocity and with image, audio, and video coming later. Try it astatine nary outgo successful the Azure OpenAI Studio Playground.

We’re astir excited astir the caller lawsuit experiences that tin beryllium enhanced with GPT-4o mini, peculiarly streaming scenarios specified arsenic assistants, codification interpreter, and retrieval which volition payment from this model’s capabilities. For instance, we observed the unthinkable velocity portion investigating GPT-4o mini connected GitHub Copilot, an AI brace programmer that assists you by delivering codification completion suggestions successful the tiny pauses betwixt keystrokes, rapidly updating recommendations with each caller quality typed.

We are besides announcing updates to Azure OpenAI Service, including extending information by default for GPT-4o mini, expanded information residency, and worldwide pay-as-you-go availability, positive show upgrades. 

Azure AI brings information by default to GPT-4o mini

Safety continues to beryllium paramount to the productive usage and spot that we and our customers expect.

We’re pleased to corroborate that our Azure AI Content Safety features—including punctual shields and protected worldly detection— are present ‘on by default’ for you to usage with GPT-4o mini connected Azure OpenAI Service.

We person invested successful improving the throughput and velocity of the Azure AI Content Safety capabilities—including the instauration of an asynchronous filter—so you tin maximize the advancements successful exemplary velocity portion not compromising safety. Azure AI Content Safety is already supporting developers crossed industries to safeguard their generative AI applications, including crippled improvement (Unity), taxation filing (H&R Block), and acquisition (South Australia Department for Education).

In addition, our Customer Copyright Commitment volition use to GPT-4o mini, giving bid of caput that Microsoft volition support customers against third-party intelligence spot claims for output content.

Azure AI present offers information residency for each 27 regions

From time one, Azure OpenAI Service has been covered by Azure’s information residency commitments.

Azure AI gives customers some flexibility and power implicit wherever their information is stored and wherever their information is processed, offering a implicit information residency solution that helps customers conscionable their unsocial compliance requirements. We besides supply prime implicit the hosting operation that meets business, application, and compliance requirements. Regional pay-as-you-go and Provisioned Throughput Units (PTUs) connection power implicit some information processing and information storage.

We’re excited to stock that Azure OpenAI Service is present disposable successful 27 regions including Spain, which launched earlier this period arsenic our ninth portion successful Europe.

Azure AI announces planetary pay-as-you-go with the highest throughput limits for GPT-4o mini

GPT-4o mini is present disposable utilizing our planetary pay-as-you-go deployment astatine 15 cents per cardinal input tokens and 60 cents per cardinal output tokens, which is importantly cheaper than erstwhile frontier models.

We are pleased to denote that the planetary pay-as-you-go deployment enactment is mostly disposable this month, allowing customers to wage for the resources they consume, making it flexible for adaptable workloads, portion postulation is routed globally to supply higher throughput, and inactive offering power implicit wherever information resides astatine rest.

Additionally, we admit that 1 of the challenges customers look with caller models is not being capable to upgrade betwixt exemplary versions successful the aforesaid portion arsenic their existing deployments. Now, with planetary pay-as-you-go deployments, customers volition beryllium capable to upgrade from existing models to the latest models.

Global pay-as-you-go offers customers the highest imaginable scale, offering 15M tokens per infinitesimal (TPM) throughput for GPT-4o mini and 30M TPM throughput for GPT-4o. Azure OpenAI Service offers GPT-4o mini with 99.99% availability and the aforesaid manufacture starring velocity arsenic our spouse OpenAI.

Azure AI offers starring show and flexibility for GPT-4o mini

Azure AI is continuing to put successful driving efficiencies for AI workloads crossed Azure OpenAI Service.

GPT-4o mini comes to Azure AI with availability connected our Batch work this month. Batch delivers precocious throughput jobs with a 24-hour turnaround astatine a 50% discount complaint by utilizing off-peak capacity. This is lone imaginable due to the fact that Microsoft runs connected Azure AI, which allows america to marque off-peak capableness disposable to customers.

We are besides releasing fine-tuning for GPT-4o mini this period which allows customers to further customize the exemplary for your circumstantial usage lawsuit and script to present exceptional worth and prime astatine unprecedented speeds. Following our update past period to power to token based billing for training, we’ve reduced the hosting charges by up to 43%. Paired with our debased terms for inferencing, this makes Azure OpenAI Service fine-tuned deployments the astir cost-effective offering for customers with accumulation workloads.

With much than 53,000 customers turning to Azure AI to present breakthrough experiences astatine awesome scale, we’re excited to spot the innovation from companies similar Vodafone (customer cause solution), the University of Sydney (AI assistants), and GigXR (AI virtual patients). More than 50% of the Fortune 500 are gathering their applications with Azure OpenAI Service.

We can’t hold to spot what our customers bash with GPT-4o mini connected Azure AI!

---

¹GPT-4o mini: advancing cost-efficient quality | OpenAI

The station OpenAI’s fastest model, GPT-4o mini is present disposable connected Azure AI appeared archetypal connected Microsoft Azure Blog.