Microsoft Unveils 7 New MAI Models at Ignite 2026: Price-Performance Focus & 'Licensed Weights' for Custom Frontier AI
Microsoft unveiled seven new MAI models at its Ignite 2026 conference on November 17, 2026, emphasizing price-performance for AI agents and introducing "MAI Licensed Weights" to enable customers to customize and run frontier models in their own Azure environments. For broader context, explore our AI Tools Pricing.
Microsoft's Strategic Shift to Price-Performance
The new MAI models are strategically positioned to deliver strong performance at competitive price points, particularly for always-on AI agents. This approach contrasts with a singular focus on achieving peak frontier capabilities, instead prioritizing efficiency and cost-effectiveness for practical enterprise applications.
Introducing MAI Licensed Weights
A key announcement at Ignite 2026 was the introduction of "MAI Licensed Weights." This initiative grants trusted customers access to Microsoft's frontier models, enabling them to customize these models and run them within their dedicated Azure infrastructure. This move aims to provide greater flexibility and control over advanced AI deployments.
New MAI Models and Their Benchmarks
The seven new MAI models span various domains, including coding, cybersecurity, image editing, and real-time speech processing. Each model is designed with specific performance and cost advantages:
- MAI Code 2: This coding model achieves 96% of the performance of Claude Opus 5.5 while operating at 5% of its price. For GitHub Copilot users, MAI Code 2 is available at a promotional price until May 17, 2027, costing 0.50 USD per million input tokens and 2.50 USD per million output tokens.
- MAI Cyber 2 Flash: In the CyberGym-E2E security benchmark, MAI Cyber 2 Flash surpassed GPT-6 Astra, delivering its performance at under 2% of the cost per task.
- MAI Image 2.7: Ranked #1 in image editing by Artificial Analysis, MAI Image 2.7 is priced at one-fifth the cost of GPT Image 2.6 Sunburst (max).
- MAI Realtime 1: Microsoft's inaugural live speech-to-speech model, MAI Realtime 1, is available starting at 0.01 USD per audio minute.
- Decision 1.1: This model leads the Artificial Analysis Decision Index, outperforming Jev, Google, and OpenAI. It is 50 times faster and 79 times cheaper than GPT-6.1 Sol.
Maia 200 AI Accelerator in Production
Alongside the new MAI models, Microsoft confirmed that its Maia 200 AI accelerator is now in production. This accelerator is engineered to enhance efficiency, delivering 1.7 times more tokens per watt, which supports the cost-performance focus of the new MAI model family.
Conclusion
Microsoft's Ignite 2026 announcements signal a strategic emphasis on delivering high-value AI solutions through a combination of optimized models and flexible deployment options. The introduction of MAI Licensed Weights and the focus on price-performance across a diverse set of new MAI models aim to provide enterprises with more efficient and customizable AI capabilities for their specific operational needs.
Sources
- AI alone won't change your business. The system running it will. - The Official Microsoft Blog
- Ignite 2026: Building the frontier ecosystem - The Official Microsoft Blog
- MAI-Code-1-Flash is now available for GitHub Copilot · community · Discussion #197306 · GitHub
- the-ledger/content/events/2026/65-microsoft-mai-model-family-build-2026.md at 6bfc3864a43a0778865baf9d72623aac0081f2f1 · pedro-bright/the-ledger · GitHub
- Microsoft Build 2026: Be yourself at work - The Official Microsoft Blog
Recommended AI tools
Perplexity
Search & Discovery
Clear answers from reliable sources, powered by AI.
Cursor
Code Assistance
Your coding agent for building ambitious software
Gemini Enterprise Agent Platform
Data Analytics
Google Cloud’s enterprise platform for building and running AI agents — formerly Vertex AI.
Adobe Firefly
Image Generation
Create your way with Adobe Firefly—AI for every creative vision.
Google AI Studio
Productivity & Collaboration
The fastest way to build AI-first applications with Google Gemini.
Hugging Face
Scientific Research
Democratizing good machine learning, one commit at a time.
About the Author

Albert Schaper is a co-founder of Best-AI.org. He focuses on product strategy, AI adoption, practical tool selection, and educational content that helps users compare AI products with clearer context.
More from AlbertWas this article helpful?
Found outdated info or have suggestions? Send us a note.