Azure OpenAI Pricing (November 2026): GPT-6 Costs, PTU and Batch

Last Updated: November 6, 2026

Summary: Azure OpenAI pricing in November 2026 runs from $0.10 per million input tokens and $0.50 per million output tokens for GPT-6 Luna up to $10 and $50 for GPT-6 Astra on Global Standard deployments. GPT-6.1 Sol costs $2 and $10. Those Global rates match OpenAI's own API. Keeping data in the US adds 10%, the EU and Asia-Pacific add 20%, prompts over 272,000 tokens cost more, and Batch cuts the bill in half.

Azure OpenAI is the way many large companies buy GPT models. It runs OpenAI's models inside Microsoft's cloud, under your existing Azure contract, security controls and data residency rules. The catch is that the pricing has more moving parts than OpenAI's own price list: deployment types, data zones, long-context tiers and reserved capacity all change what you pay.

This guide breaks down Azure OpenAI pricing for every current GPT-6 model, explains each deployment type, compares Azure with buying direct from OpenAI, and works through real monthly cost examples. Prices come from Microsoft's own Foundry announcements and pricing pages, checked in November 2026. All figures are in US dollars per million tokens.

Azure OpenAI charges per token, with no monthly fee for pay-as-you-go use. On Global Standard deployments, the four current GPT-6 family models cost between $0.10 and $10 per million input tokens and between $0.50 and $50 per million output tokens. Output costs five times as much as input on every one of them, so long answers drive most of the bill.

Prices are Global Standard, short-context rates from Microsoft's GPT-6 Foundry announcement (September 22, 2026) and its GPT-6.1 Sol announcement (September 29, 2026). Writing a prompt into the cache costs 1.25 times the normal input rate, for example $12.50 per million tokens on Astra and $2.50 on Sol.

All of these models share a 1,050,000-token context window with up to 128,000 output tokens, according to Microsoft's Foundry model documentation. Microsoft also lists GPT-6.1 Terra (version 2026-10-26) in the catalog. OpenAI charges $1 and $5 for Terra on its own API, but we could not confirm an Azure rate for it, so check the Azure pricing calculator before you budget for it.

The deployment type you choose decides where your data is processed, how fast you get answers and what you pay. This is the part of Azure OpenAI pricing that confuses most first-time buyers.

The Data Zone premiums come straight from Microsoft's GPT-6.1 Sol announcement, which says US Data Zone pricing is a 10% premium over Global and EU and APAC are a 20% premium. Azure's pricing page says Batch returns completions within 24 hours "for a 50% discount on Global Standard Pricing."

If your company has to keep customer data in the EU for GDPR or contract reasons, budget 20% more than the headline price. For most US companies without a residency requirement, Global Standard is the cheapest real-time option.

Azure OpenAI bills short-context and long-context requests at different rates. Microsoft's documentation says the input token count decides which category a request falls into, and Azure price listings compiled by AzureSpeed put the cutoff at 272,000 input tokens, the same threshold OpenAI uses on its own API.

In other words, a very long prompt doubles the input rate and raises the output rate by 50%. If you feed whole contracts, codebases or document sets into a single request, this tier will often decide your bill. Splitting work into smaller requests or using retrieval to send only relevant passages keeps you on the cheaper rate.

Provisioned Throughput Units let you reserve a fixed amount of model capacity instead of paying per token. You pay for the capacity whether or not you use it, so PTUs make sense only for steady, high-volume production traffic where predictable speed matters more than flexibility.

Azure's pricing page lists these minimum purchases:

At launch, PTUs were available for GPT-6 Astra and Sol across Global regions and the US and EU Data Zones, and for GPT-6.1 Sol through Global and US Data Zone deployments. Microsoft says provisioned pricing "varies by deployment type" and does not publish per-PTU figures in its model announcements. Use the Azure pricing calculator, signed in to your own account, to see the hourly and reservation rates under your agreement.

On list price, neither. Azure's Global Standard rates for GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna match what OpenAI charges on its own API, and both double the input price for prompts over 272,000 tokens. The real differences are in billing, data controls and the extras around the model.

OpenAI direct rates come from our OpenAI API pricing guide, which was checked against OpenAI's developer pricing page on November 1, 2026.

Pick Azure if your company already spends on Microsoft's cloud, needs Azure security and compliance controls, or wants AI costs to count toward an existing Azure commitment. Pick OpenAI direct if you are a startup or small team that wants OpenAI's own tools and the least setup.

Here is what three common workloads would cost per month on pay-as-you-go pricing, using the published rates above.

500 million input tokens and 100 million output tokens a month, Global Standard: (500 x $0.10) + (100 x $0.50) = $100 a month. The same traffic in an EU Data Zone costs (500 x $0.12) + (100 x $0.60) = $120.

200 million input tokens, half of them cache hits, and 40 million output tokens a month, all short-context:

Without caching, the Global bill would be $800, so caching saves $190 a month on this workload alone.

20 million input tokens and 5 million output tokens, short-context, Global Standard: (20 x $10) + (5 x $50) = $450 a month. Run the same job through Batch and it drops to $225.

These examples cover model tokens only. Built-in tools such as file search and code interpreter, vector storage and networking are billed separately.

There is no permanently free Azure OpenAI tier. New Azure accounts get a $200 credit to use within 30 days, according to Microsoft's pricing page, and that credit can go toward Azure OpenAI usage. Vector storage for file search includes 1 GB free. After that, every token is billed.

Access is also tiered by quota. Microsoft's documentation says some quota tiers need a quota request before deploying GPT-6 models, while Tier 5 and Tier 6 subscriptions have quota by default.

How much does Azure OpenAI cost per token?
On Global Standard deployments, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, GPT-6.1 Sol costs $2 and $10, and GPT-6 Astra costs $10 and $50. US Data Zone processing adds 10%, EU and APAC add 20%, and prompts over 272,000 tokens are billed at higher long-context rates.

Is Azure OpenAI cheaper than the OpenAI API?
No. Azure's Global Standard prices for GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna match OpenAI's own API rates, and both offer a 50% Batch discount. Azure can work out cheaper in practice if usage counts toward an existing Microsoft commitment, and it also sells reserved capacity through Provisioned Throughput Units.

What is a PTU in Azure OpenAI?
A Provisioned Throughput Unit is a block of reserved model capacity that you pay for by the hour or through a monthly or yearly reservation, instead of paying per token. Global and Data Zone deployments start at 15 PTUs for most models. PTUs suit steady, high-volume production workloads that need predictable speed.

Does Azure OpenAI have a free tier?
No. There is no permanent free tier, but new Azure accounts receive a $200 credit for 30 days that can be used on Azure OpenAI. File search vector storage includes 1 GB free. After the credit runs out, usage is billed per token or per PTU.

What is the cheapest model on Azure OpenAI?
GPT-6 Luna is the cheapest current GPT-6 family model, at $0.10 per million input tokens and $0.50 per million output tokens on Global Standard. Batch processing halves that again. Microsoft positions Luna for high-volume work such as extraction, summarization, request routing and routine customer conversations.

Why is the Azure OpenAI pricing page showing no prices?
The public Azure OpenAI pricing page sometimes shows placeholder dashes instead of numbers. Use the Azure pricing calculator or sign in to your Azure account to see the rates that apply to your agreement, including any negotiated discounts.

Azure OpenAI pricing starts at the same list price as OpenAI's API, then shifts with every choice you make: where data is processed, how long your prompts are, whether you can wait for Batch, and whether you reserve capacity. For most teams the formula is simple. Default to Global Standard, cache repeated prompts, send background jobs to Batch, and use Luna for anything that doesn't need Sol or Astra.

If you already run on Microsoft's cloud, Azure is usually the easier way to buy GPT models at scale. If you don't, compare it against Claude API pricing and Gemini API pricing before you commit.

Written by the AI Business Weekly Editorial Team. Reviewed, edited and published by Sameer Khan. This article was AI-assisted, then reviewed by Sameer Khan before publishing.