Microsoft launches MAI Licensed Weights for customizable, self-hosted AI models
Microsoft introduced MAI Licensed Weights, letting businesses take frontier MAI models, customize them with proprietary data, and run them within their own Azure environments under their control. The company cited AT&T's custom MAI model for 5G network analytics as outperforming GPT-6.1 Sol at 50 times lower cost, and said its Code 2 model delivers frontier-level coding performance at lower prices than competitors.
GoKawiil's interpretation of the reporting above, not reported fact.
By offering licensed weights that companies can host and adapt themselves, Microsoft is addressing enterprise concerns about data ownership and trust in third-party models, which could help it compete against open-weight and rival proprietary AI offerings. The emphasis on price-performance and customer control suggests Microsoft is positioning its MAI model family as a differentiated option for businesses wary of ceding intellectual property to external AI providers.
- Microsoft introduced MAI Licensed Weights, allowing customers to run customized frontier models in their own Azure environment.
- AT&T's custom MAI model reportedly outperformed GPT-6.1 Sol at 50x lower cost for 5G analytics.
- Microsoft's MAI family includes models for image editing, transcription, cyber defense, and coding, emphasizing affordability and customer data control.
Source: microsoft.ai, 2026-11-18
Published there as: “Microsoft AI: Seven new world-class models at scalable price-performance”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.