Microsoft Artificial Intelligence launches seven new MAI models

Microsoft Artificial Intelligence announced seven new in-house MAI models spanning reasoning, coding, image, transcription, and voice. The company is pairing the launch with Frontier Tuning, a workflow-specific adaptation approach for enterprise-controlled models.

Microsoft Artificial Intelligence announced a family of seven new models developed in-house and framed the launch as part of a broader effort to build a superintelligence lab. The compute used to train frontier models has increased by a factor of one trillion. Microsoft now expects another thousand-fold increase over the next three years, which in turn means more advanced capabilities, and the continued rollout of ever more effective Artificial Intelligence. The company says that ramp will change work, business, and daily life, and positions MAI as a system for keeping users and organizations close to the frontier as capabilities improve.

The new MAI model family spans image, voice, transcription, coding, and reasoning. MAI-Thinking-1 is Microsoft Artificial Intelligence’s flagship reasoning model, a medium-sized system trained from the ground up on clean data, without distillation from third-party models. MAI-Code-1-Flash is an inference-efficient agentic coding model. This model is tailor-made for and deeply integrated into GitHub Copilot, VS Code and the Microsoft stack, and, with 5 billion active parameters, is comparable to Haiku but cheaper. MAI-Image-2.5 including its ultra-efficient Flash variant, supports both world-class text-to-image and image editing, surpassing the Arena score of Nano Banana Pro. MAI Transcribe-1.5 is described as the best transcription model in the world, with SOTA accuracy. It’s five times faster than competing models, with built-in support for domain-specific terminology across 43 languages. MAI-Voice-2 brings high-quality, natural-sounding speech generation across 15 languages, with the ability to adapt to a voice from a short sample, alongside strong safeguards against misuse.

Microsoft is also introducing Microsoft Frontier Tuning, a reinforcement learning approach intended to adapt models to specific workflows inside an organization. Reinforcement learning environments allow MAI models to learn directly from workflow traces, including steps, decisions, and actions, while keeping the adaptation controlled by the customer. Across Microsoft and with customers, Frontier Tuning is showing that custom models are both better and more efficient: Microsoft’s MAI tuned model for Excel matches GPT 5.4 while being up to 10× more efficient. When tuned for a market-leading organization’s exacting enterprise standards, MAI achieved the highest win rate of any model tested at roughly 10× lower cost.

Microsoft and Mayo Clinic are collaborating to co-create a frontier Artificial Intelligence model for healthcare, combining Mayo Clinic’s clinical expertise, de-identified clinical data, and longitudinal insights with Microsoft’s foundational capabilities. The model will be deployed within Mayo Clinic’s environment before broader availability through Microsoft Foundry once validated, and Mayo Clinic will own the model. Microsoft says its lab trains reasoning models from scratch, avoids unlicensed or opaque data, and builds each component internally. Microsoft co-designs with its own Maia 200 silicon, and is already seeing a 1.4x efficiency boost from these efforts. The long-term goal is Humanist Superintelligence, with advanced systems remaining tools shaped by human intent and oversight.

78

Impact Score

Artificial Intelligence designed vaccine targets coronavirus threats

University of Cambridge researchers have trialled a vaccine component designed entirely by Artificial Intelligence to train the immune system against a broad family of coronaviruses. Early safety work showed modest immune effects, while larger studies and related vaccine projects are underway.

Meta Instagram breach exposes Artificial Intelligence agent security gaps

Attackers exploited Meta’s Artificial Intelligence customer support agent to take over Instagram accounts, underscoring risks that go beyond advanced hacking models. Security researchers warn that agentic systems can create serious vulnerabilities when deployed without strong guardrails and red-teaming.

Broadcom falls on softer Artificial Intelligence chip outlook

Broadcom’s Artificial Intelligence chip outlook overshadowed an earnings beat, pressuring Advanced Micro Devices and Intel as investors reassessed semiconductor momentum. The selloff reflected high expectations after a sharp run in chip stocks.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.