From the source
Microsoft AI used its Ignite 2026 keynote to introduce MAI Licensed Weights, which lets customers customize a frontier MAI model with their own data and run it inside their own Azure environment.
- The company says CIOs and CFOs tell it that token spend is going through the roof and that frontier prices make the math unsustainable, so it co-optimizes quality and price rather than chasing frontier quality alone.
- On coding, MAI says MAI Code 2 delivers 98% of GPT 6.1 Sol's performance on four enterprise coding benchmarks at 40% of the cost, and 5% of Opus 5.5's cost, while Code 2 Flash matches 6 Luna and Haiku 5.5 at 65% and 20% of their prices.
- Decision 1.1 adds model routing inside applications and beats Jev, Google and
OpenAI on the AA Decision Index, according to the company.
- MAI says the models are built for everyday coding and are explicitly not yet tuned for harder problems such as fundamental science research and very long horizon agentic work.
- MAI Cyber 2 Flash is ahead of GPT 6 Astra on Cyber Gym end-to-end at less than 2% of the cost per task, and MAI Image 2.7 is described as the world's number one image editing model at a fifth the cost of the competition.
- MAI Transcribe has been deployed to 10 million patient Dragon Copilot encounters a month while halving the error rate, and
Microsoft says customers including Adobe, Sports Illustrated and
Manus use the image model for precise edits.





