--- title: Microsoft launches MAI-Image-2.5-Pro and... description: Microsoft has launched two new proprietary AI models in public preview: MAI-Image-2.5-Pro, its highest-fidelity image generation model priced at $106 per 1M... canonical: https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash, its own image and voice AI models | daily.dev og:description: Microsoft has launched two new proprietary AI models in public preview: MAI-Image-2.5-Pro, its highest-fidelity image generation model priced at $106 per 1M... og:url: https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao og:image: https://api.daily.dev/og/posts/ZDpQLsfAO.png og:image:alt: Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash, its own image and voice AI models og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash, its own image and voice AI models **[Collections](https://daily.dev/sources/collections)** · 3 min read · 3 upvotes · 0 comments ## Summary Microsoft has launched two new proprietary AI models in public preview: MAI-Image-2.5-Pro, its highest-fidelity image generation model priced at $106 per 1M image output tokens, and MAI-Voice-2-Flash, a voice model that is twice as fast and 32% cheaper than its predecessor at $15 per 1M characters. Both models are being integrated into Microsoft products like Bing, PowerPoint, and OneDrive. The move signals Microsoft's strategy to reduce reliance on external AI providers, including OpenAI, by building more of its AI stack in-house. ## Content ## What launched Microsoft quietly dropped two new models into public preview this week. **MAI-Image-2.5 Pro** is their highest-fidelity image generation model so far. Pricing: $5 per 1M text input tokens, $8 per 1M image input tokens, and $106 per 1M image output tokens. **MAI-Voice-2 Flash** is a low-latency voice model — 2x faster than MAI-Voice-2 and 32% cheaper, at $15 per 1M characters. Both are available through Azure and are already being deployed across Bing, PowerPoint, and OneDrive, replacing the external models that previously handled those workloads. --- ## The broader strategy These aren't one-off releases. MAI models now power more than half of all Microsoft products, and the pattern is consistent: build specialized models trained directly inside specific products using reinforcement learning and real customer workflows, then route traffic to them whenever they match or beat the alternatives on quality. The Excel MAI model is reportedly delivering quality on par with GPT-4.5 for the most common spreadsheet tasks, at significantly lower cost. In GitHub Copilot, MAI-Code-1-Flash is outperforming similarly sized models on code acceptance rates while using fewer tokens per request. The image models replacing OpenAI's in PowerPoint and Bing run about 85% cheaper. Microsoft's framing is that the model itself is only one piece of the system. Memory, context, tools, agent orchestration, and product-specific evaluations all matter. By optimizing the whole stack together rather than just swapping in a better base model, they can keep improving quality while bringing costs down. Frontier models from OpenAI and Anthropic stay in the mix for tasks that genuinely need them. But for everyday workloads, traffic increasingly goes to MAI. --- ## Why eval-driven routing matters The interesting thing here isn't just that Microsoft built cheaper models — it's the routing discipline behind the deployment. The approach is: measure quality per task, set a quality floor, and send requests to the cheapest model that clears it. That's a meaningful shift from the default of picking one frontier model and sending everything to it. This is becoming a real pattern. Kilo's Auto Efficient router works similarly — classify task difficulty, route to the cheapest sufficiently accurate model — and reports 71% of frontier quality at 72% lower cost. As AI token spend becomes an actual budget line item rather than a rounding error, multi-model routing stops being an optimization trick and starts being the obvious default. Microsoft is just doing it at a scale that makes the signal hard to ignore. --- ## Other Azure updates this week A few other things worth noting from the July 24 Azure update: - TLS encryption in transit for AKS Azure Files NFS mounts - PowerShell 7.6 and Python 3.14 support in Azure Functions - IPv6 GA support for Azure VPN Gateway - Standard Service Endpoints preview for scalable PaaS access control across multiple subnets - Custom DDoS protection policy thresholds in preview - Defender security assessments GA for PostgreSQL Flexible Server - A dedicated generative AI logging table in Log Analytics --- ## One more thing: SMS logins are going away Unrelated to MAI but worth flagging — Microsoft is retiring SMS-based authentication. Passkeys become the default for Entra ID on September 1, and SMS/voice codes will be fully discontinued by February 2027. The stated reason is AI-powered phishing, which apparently achieves a 54% click-through rate compared to 12% for conventional phishing. SMS codes are too easy to intercept in that environment to keep around. --- Tags: [#microsoft](https://daily.dev/tags/microsoft), [#genai](https://daily.dev/tags/genai), [#openai](https://daily.dev/tags/openai), [#text-to-speech](https://daily.dev/tags/text-to-speech) [View this post on daily.dev](https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash, its own image and voice AI models","url":"https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao"},"datePublished":"2026-07-24T09:22:53.147Z","dateModified":"2026-07-28T02:03:24.066Z","description":"Microsoft has launched two new proprietary AI models in public preview: MAI-Image-2.5-Pro, its highest-fidelity image generation model priced at $106 per 1M...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/microsoft-launches-mai-image-2-5-pro-and-mai-voice-2-flash-its-own-image-and-voice-ai-models-zdpqlsfao","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":3},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"microsoft,genai,openai,text-to-speech","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"Microsoft launches MAI-Image-2.5-Pro and MAI-Voice-2-Flash, its own image and voice AI models"}]} ```