Microsoft AI Models Promise Up to 89% Lower Costs

Key takeaways

• Microsoft says its new in-house models can cost up to 89% less than OpenAI choices.
• The models give firms another option inside Azure AI Foundry.
• Lower prices matter most for apps that handle millions of requests.
• Companies should test answer quality, speed, and safety before switching.

Microsoft AI models are the company’s own systems for creating text, code, and answers. Microsoft says its new releases can cost up to 89% less than comparable OpenAI options. That could help firms run AI tools for far less money, but they should test quality first.

What are Microsoft AI models and why did Microsoft launch them?

Microsoft has launched new in-house AI options for business customers using Azure AI Foundry. Azure AI Foundry is Microsoft’s platform for building and running AI apps. The move gives customers more choices beyond models from OpenAI and other outside firms.

For years, Microsoft has invested heavily in OpenAI and sold OpenAI models through Azure. That relationship remains important. But building its own models lets Microsoft control more of the tech, the price, and the way it fits into its cloud service.

The new Microsoft AI models target a simple problem: running AI can get costly fast. A chatbot may answer thousands of questions each day. A coding tool at a large company may handle far more.

Microsoft’s 89% figure is a price claim against selected OpenAI alternatives. It does not mean every task will cost 89% less. The final bill depends on the model, the task, and how much text the app sends and receives.

How does the 89% cost claim work?

AI providers usually charge by tokens. Tokens are small chunks of text that a model reads or writes. A short word may be one token, while a long word can use several.

Microsoft says its models can reduce costs by as much as 89% in certain comparisons. Put simply, a job that costs $100 under one option could cost about $11 under that claim. That is a large gap for a company running the same job again and again.

Illustrative cost for the same workloadComparison option: $100Microsoft claim: about $11Maximum claimed saving: 89%

The chart is only an easy example, not a fixed price list. Input text and output text often have different rates. Long answers can raise costs because the model creates more tokens.

What to compare	Why it matters
Listed model price	It sets the starting cost per token.
Input and output length	Long prompts and answers increase the bill.
Quality on your work	A cheap model is not useful if it makes more errors.
Speed and limits	Slow replies or request caps can affect an app.

Buyers can check current rates in Microsoft’s Azure AI Foundry pricing pages. Prices and model choices can change, so a sales slide should never be the only thing a team uses.

Where could Microsoft AI models save the most money?

The biggest gains may come from high-volume, repeat jobs. Think of a shop that sorts 50,000 customer messages a day. Or picture a bank that turns long call notes into short summaries.

Those jobs do not always need the strongest model available. A smaller or cheaper model may handle routine work well enough. Then a firm can send only tricky questions to a more powerful, pricier model.

This approach is called model routing. Model routing means sending each job to the model that best fits it. It can cut spending, but it needs careful rules and regular checks.

India’s growing data-centre push also makes cloud AI costs a bigger business issue. HCLTech’s planned ₹730 crore Odisha data centre shows how companies are spending to support more digital workloads. Lower model prices could make some AI projects easier to approve.

Will cheaper Microsoft AI models hurt OpenAI?

Not by themselves. OpenAI still offers widely used models, and many firms value their abilities for hard reasoning, writing, and coding tasks. Microsoft can also continue selling OpenAI models through its cloud platform.

Still, the launch adds pressure in a crowded market. Google, Anthropic, Meta, Amazon, and many smaller firms also want business AI customers. Price now sits beside quality as a major reason to choose a model.

For Microsoft, having in-house options reduces its dependence on one supplier. It also helps the company offer a wider menu to customers. That matters while businesses try to control AI budgets.

The same budget pressure is visible across technology services. Infosys cutting its FY27 revenue forecast reflects cautious global IT spending. Firms may welcome tools that promise lower running costs.

What should a company test before changing models?

First, use real company tasks, not just a public benchmark. A benchmark is a standard test used to compare systems. It can help, but it may not match a firm’s own work.

Next, measure accuracy on at least 100 to 500 sample tasks. Check whether the model gets facts right. Also check if it follows instructions and knows when it is unsure.

Security matters as much as price. Teams should know where their data goes and who can access it. Microsoft explains its service controls in its Azure AI Foundry model catalog documentation.

Finally, compare the total cost. Include engineering time, safety checks, and any errors people must fix. The cheapest price per token is not always the lowest cost in real life.

Why this launch matters

Microsoft AI models could make basic AI features cheaper for many businesses. Microsoft’s headline claim is a maximum saving of 89%, not a promise for every customer. The useful question is whether a model delivers good enough results at a lower total cost.

That is why the launch is bigger than a price cut. It gives companies another way to build chatbots, search tools, and coding helpers. As more models compete, buyers may gain more control over both their AI bills and their choices.

FAQs

How much cheaper are Microsoft AI models?

Microsoft says selected new models can cost up to 89% less than comparable OpenAI options. Actual savings depend on the model and the amount of text processed.

What is Azure AI Foundry?

Azure AI Foundry is Microsoft’s service for building, testing, and running AI applications. It offers a catalog of models from Microsoft and other providers.

Why should firms test a cheaper AI model?

A lower price does not guarantee good answers. Firms should test quality, speed, data controls, and the cost of fixing mistakes before a full switch.