EXPLORINEER
en

DE — heise online

Microsoft unveils new AI models and expands its licensing business

AI1 min read

Microsoft has shown seven new in-house AI models at its Ignite conference. It wants them to match rival models in their areas while costing less, and it plans to license its model weights to selected business customers.

Seven new AI models

At its Ignite conference, Microsoft showed seven new AI models. They cover coding, cybersecurity, image creation, voice chat, speech-to-text, and decision-making. Microsoft says each model can keep up with leading rival models in its area, but at a lower cost. The company is competing more and more with OpenAI and Anthropic.

Key numbers

  • MAI Code 2 reaches 98 percent of the performance of GPT-6.1 Sol, averaged over four benchmarks (tests that score AI models).
  • MAI Code 2 costs about 20 percent less, without the action discount.
  • MAI Cyber 2 Flash performs about as well as GPT-6 Astra on a security-flaw test. Microsoft says it costs more than 98 percent less per test task.
  • Both MAI Code 2 models already run in GitHub Copilot.

Renting models through a license

Until now, many companies used Microsoft's AI models only through Microsoft's own products and cloud service. A new license program will let selected companies get direct access to the model weights. The weights are the numbers that make an AI model work. With them, a company can adjust the model and run it in its own Azure cloud setup. Microsoft is not making the weights public. It already works with AT&T, Toyota, and Alaska Airlines on custom models.

Why this matters

If Microsoft's own models work well and cost less, the company may rely less on OpenAI and other AI makers. Because the models still run on Azure, Microsoft could also grow its cloud business. Microsoft also says it wants to build a top general-purpose model in the long run, but it has not named a release date.

What to watch

  • Which companies join the new license program.
  • Whether the cost savings hold up in real use outside Microsoft's tests.
  • When Microsoft releases its planned top general-purpose model.

Glossary

Ignite
Microsoft's yearly conference for businesses and IT workers, where the company shows new products.
OpenAI
An AI company that makes the GPT models and the ChatGPT chatbot.
Anthropic
An AI company that makes the Claude family of AI models.
GitHub Copilot
A tool that uses AI to help programmers write code.
Azure
Microsoft's cloud computing service, where companies run software and data on rented servers.

Same story, different eyes

Ad

Add 150 words. Learn 10 a day.

Dicter word books send only the words you learn to your review notes.

AI전자신문

Lablup Posts 48% Year-on-Year Revenue Growth for Nine Months, First Earnings Report Since Listing

Lablup reported cumulative consolidated revenue of 6.99 billion won and operating profit of 900 million won for the first three quarters, up 48.2% and 36.5% year on year, driven by corporate demand for AI infrastructure. It posted a cumulative net loss of 2.4 billion won due to a 3.49 billion won derivative liability valuation loss tied to the conversion of RCPS shares, but returned to a net profit of 360 million won in the third quarter as that accounting effect expired.

AIZDNet Korea

Maum AI Signs Physical AI Industry-Academia MOU with Hanyang University ERICA and Doosan Robotics

Maum AI signed a three-way MOU with Hanyang University ERICA and Doosan Robotics at the "AI Symposia" held on the 18th at Hanyang University ERICA to build cooperation on physical AI talent development and industry-academia collaboration. The partnership is based on the AX Graduate School project, a government-backed AI talent program run by the Ministry of Science and ICT, with cooperation planned in joint curricula, industry projects, internships, joint R&D and demonstrations, and government R&D planning. Maum AI will contribute its generative and multimodal AI and AI platform capabilities to the effort.

AIheise online

Google Genie 4: Interactive AI Worlds Now Available to Subscribers

Google is rolling out its Genie 4 AI world model to Google AI Ultra subscribers via the Project Genie web app, with access also available in Germany. Users of the cheaper Google AI Pro plan can join a waitlist, and daily usage limits apply. Genie 4 generates interactive 3D environments from text or images, and the model currently supports only about one minute of continuous interaction.

AIheise online

Gemini 4 Neon Challenges Google's Flagship Model

Google DeepMind has released Gemini 4 Neon, the fourth model in the Gemini 4 family within seven weeks, designed for efficiency while matching and slightly outperforming flagship model Carbon on some benchmarks. Neon is slated to become cheaper per token than Carbon from December, when Carbon's introductory prices expire. Google positions it for complex programming, analytical reasoning and enterprise use, filling the gap between Carbon and the everyday model Hydrogen.