Azure OpenAI Service pricing overview
Unlock the power of Azure OpenAI Service's generative AI models with flexible Standard (On-Demand) and Provisioned Throughput Units (PTUs). The Standard model lets you pay only for tokens processed, while PTUs ensure consistent throughput and minimal latency variance for scalable solutions. Pricing includes costs per 1,000 tokens, and PTU rates provide a predictable cost structure. Azure OpenAI Service offers advanced capabilities like GPT-4o, fine-tuning for customization, DALL-E for image generation, and Whisper for speech-to-text. For personalized guidance on optimizing AI deployments, contact a sales specialist.
Explore pricing options
Apply filters to customise pricing options to your needs.
Prices are estimates only and are not intended as actual price quotes. Actual pricing may vary depending on the type of agreement entered with Microsoft, date of purchase, and the currency exchange rate. Prices are calculated based on US dollars and converted using London closing spot rates that are captured in the two business days prior to the last business day of the previous month end. If the two business days prior to the end of the month autumn on a bank holiday in major markets, the rate setting day is generally the day immediately preceding the two business days. This rate applies to all transactions during the forthcoming month. Sign in to the Azure pricing calculator to see pricing based on your current programme/offer with Microsoft. Contact an Azure sales specialist for more information on pricing or to request a price quote. See frequently asked questions about Azure pricing.
US government entities are eligible to purchase Azure Government services from a licensing solution provider with no upfront financial commitment, or directly through a pay-as-you-go online subscription.
Important—The price in R$ is merely a reference; this is an international transaction and the final price is subject to exchange rates and the inclusion of IOF taxes. An eNF will not be issued.
US government entities are eligible to purchase Azure Government services from a licensing solution provider with no upfront financial commitment, or directly through a pay-as-you-go online subscription.
Important—The price in R$ is merely a reference; this is an international transaction and the final price is subject to exchange rates and the inclusion of IOF taxes. An eNF will not be issued.
Pricing details:
Language models
Models | Context | Input (Per 1,000 tokens) | Output (Per 1,000 tokens) | Price per PTU per Hour | Minimum Scaling Increment | Monthly Reservation per PTU | Yearly Reservation per PTU |
---|---|---|---|---|---|---|---|
GPT-4o Global Deployment | 128K | $- | $- | N/A | N/A | N/A | N/A |
GPT-4o Regional API | 128K | $- | $- | $- | 50 PTUs | $- | $- |
GPT-4o-mini Global Deployment | 128K | $- | $- | N/A | N/A | N/A | N/A |
GPT-4o-mini Regional API | 128K | $- | $- | $- | 25 PTUs | $- | $- |
GPT-3.5-Turbo-0125 | 16K | $- | $- | $- | 100 PTUs | $- | $- |
GPT-3.5-Turbo-Instruct | 4K | $- | $- | N/A | N/A | N/A | N/A |
GPT-4-Turbo | 128K | $- | $- | $- | 100 PTUs | $- | $- |
GPT-4-Turbo-Vision | 128K | $- | $- | N/A | N/A | N/A | N/A |
GPT-4 | 8K | $- | $- | $- | 50 PTUs | $- | $- |
GPT-4 | 32K | $- | $- | $- | 200 PTUs | $- | $- |
This table provides a detailed comparison of Standard (On-Demand) versus Provisioned (PTU) pricing for various language models. The 'Context' column specifies the maximum number of tokens each model can handle per response. Pricing details for input and output tokens are listed, reflecting the cost per 1,000 tokens. The PTU pricing model includes an hourly rate and a minimum scaling increment, representing the minimum number of PTUs required for each model. The "Monthly Reservation per PTU" and "Yearly Reservation per PTU" columns indicate the reservation costs per PTU. This comparison helps users understand the cost implications of using each model under both Standard (On-Demand) and Provisioned (PTU) billing options, allowing for informed decisions based on their specific usage needs.
Language models are also now available in the Batch API that returns completions within 24 hours for a 50% discount.
Legacy Language Models
Models | Context | Input (Per 1,000 tokens) | Output (Per 1,000 tokens) |
---|---|---|---|
GPT-3.5-Turbo-0301 | 4K | $- | $- |
GPT-3.5-Turbo-0613 | 4K | $- | $- |
GPT-3.5-Turbo-0613 | 16K | $- | $- |
GPT-3.5-Turbo-1106 | 16K | $- | $- |
Assistants API
The Assistants API and its tools make it easy for developers to build AI Assistants in their applications.
The tokens used for the Assistants API are billed at the chosen language model's per token input/output rates used with each Assistant. Additionally, we charge the following fees for tool usage:
Tool | Input |
---|---|
File Search* | $-/GB of vector-storage per day (1 GB free) |
Code Interpreter** | $-/session |
*GB refers to binary gigabytes, where 1 gb is 2^30 bytes.
**If your assistant calls Code Interpreter simultaneously in two different threads, this would create two Code Interpreter sessions (2 * $-). Each session is active by default for one hour, which means that you would only pay this fee once if your user keeps giving instructions to Code Interpreter in the same thread for up to one hour.
Inference cost (input and output) varies based on the GPT model used with each Assistant. If your assistant calls Code Interpreter simultaneously in two different threads, this would create two Code Interpreter sessions (2 * $-). Each session is active by default for one hour, which means that the price is for up to one hour of giving instructions to Code Interpreter in the same thread.
Base models
Models | Usage per 1,000 tokens |
---|---|
Babbage-002 | $- |
Davinci-002 | $- |
Fine-tuning models
Models | Training per 1,000 tokens | Hosting per hour | Input Usage per 1,000 tokens | Output Usage per 1,000 tokens |
---|---|---|---|---|
Babbage-002 | $- | $- | $- | $- |
Davinci-002 | $- | $- | $- | $- |
GPT-3.5-Turbo (4K) | $- | $- | $- | $- |
GPT-3.5-Turbo (16K) | $- | $- | $- | $- |
Image models
Models | Quality | Resolution | Price (per 100 images) |
---|---|---|---|
Dall-E-3 | Standard | 1024 * 1024 | $- |
Standard | 1024 * 1792, 1792 * 1024 |
$- | |
Dall-E-3 | HD | 1024 * 1024 | $- |
HD | 1024 * 1792, 1792 * 1024 |
$- | |
Dall-E-2 | Standard | 1024 * 1024 | $- |
Embedding models
Models | Per 1,000 tokens |
---|---|
Ada | $- |
text-embedding-3-large | $- |
text-embedding-3-small | $- |
Speech Models
Models | Price |
---|---|
Whisper | $-/hour |
TTS (Text to Speech) | $-/1M characters |
TTS HD | $-/1M characters |
Azure pricing and purchasing options
Connect with us directly
Get a walkthrough of Azure pricing. Understand pricing for your cloud solution, learn about cost optimisation and request a customised proposal.
Talk to a sales specialistSee ways to purchase
Purchase Azure services through the Azure website, a Microsoft representative or an Azure partner.
Explore your optionsAdditional resources
Azure OpenAI Service
Learn more about Azure OpenAI Service features and capabilities.
Pricing calculator
Estimate your expected monthly costs for using any combination of Azure products.
SLA
Review the Service Level Agreement for Azure OpenAI Service.
Documentation
Review technical tutorials, videos, and more Azure OpenAI Service resources.
Frequently asked questions
-
Azure OpenAI Service offers pricing based on both Pay-As-You-Go and Provisioned Throughput Units (PTUs). Pay-As-You-Go allows you to pay for the resources you consume, making it flexible for variable workloads. PTUs offers a predictable pricing model where you reserve and deploy a specific amount of model processing capacity. This model is ideal for workloads with consistent or predictable usage patterns, providing stability and cost control.
-
To learn more about PTUs and Azure Open AI pricing please read PTU documentation or contact our sales specialist
Talk to a sales specialist for a walk-through of Azure pricing. Understand pricing for your cloud solution.
Get free cloud services and a $200 credit to explore Azure for 30 days.