The new family of systems targets common organizational workflows, including the MAI Code 2 pair for routine programming and MAI Cyber 2 Flash for low-cost security analysis. Capabilities extend to creative and conversational workloads with MAI Image 2.7, which leads Artificial Analysis benchmarks at a fifth of competitor costs, and MAI Realtime 1 for low-latency voice interactions. Additionally, MAI Transcribe 2 supports medical operations by handling 10 million monthly clinical encounters through Dragon Copilot. Infrastructure costs are further optimized by Decision 1.1, an intelligent routing tool that directs prompts to high-cost models only when strictly required.
Key points
Organizations can deploy persistent operational agents with significantly reduced computa…: Organizations can deploy persistent operational agents with significantly reduced computational expenses using targeted, cost-effective model architectures.
Software developers gain specialized coding and security assistants integrated directly i…: Software developers gain specialized coding and security assistants integrated directly into enterprise platforms like GitHub Copilot and Microsoft Foundry.
Operational infrastructure is streamlined by intelligent routers that prevent unnecessary…: Operational infrastructure is streamlined by intelligent routers that prevent unnecessary expenses by shielding premium models from routine business queries.
Sources checked by our analysis pipeline: 1
MAI Licensed Weights: Frontier performance you control
At MAI, our mission is to build you the very best models in the world and keep every developer and every enterprise at the frontier of capability.
But as token spend skyrockets, we know that you also want models at a price point that lets you run agents 24-7 doing real engineering work.
That’s why today we’re releasing a family of seven new world class models at a truly exceptional price-performance:
Here’s the lineup:
Security, alignment and containment are the most important priorities for agent development in 2027 and beyond. That’s our top priority. We need to run monitoring and control agents defending and patching live production systems 24-7. That’s why MAI Cyber 2 Flash was designed to deliver frontier intelligence at a fraction of the cost.
Built with Microsoft’s decades of proprietary data and world class security expertise, our models help defenders stay safe in this new era where attackers have cheap and abundant offensive tools.
MAI Cyber 2 Flash builds on earlier MAI Cyber models that defenders already rely on. MAI…
Security researchers are relying on advanced artificial intelligence models to uncover critical software flaws. These efforts identified dangerous flaws in Nginx that allow unauthorized remote code execution. The findings were independently corroborated when a separate security organization revealed it had also pinpointed the exact same security risks.
Addressing the sudden leadership vacuum at a critical hardware partner, OpenAI leaders highlighted their deep operational ties with Nvidia at the DealBook Summit. Jensen Huang's exit follows serious controversies, yet key allies are emphasizing corporate continuity to ensure their ongoing technological collaboration remains on track.
Emerging scrutiny has surfaced against the artificial intelligence developer just as it prepares for its planned initial public offering. Signatories of a newly released joint document are disputing the firm's established corporate reputation for responsibility and risk mitigation. They argue that the company's internal research agenda poses severe concerns due to its pursuit of recursive self-improvement technology.