Microsoft has introduced MAI-Image-1, its first internally developed text‑to‑image model, now integrated into Bing Image Creator and Copilot Audio Expressions. Announced in October, the model is praised for fast, photorealistic output, especially in food, nature and artistic lighting scenes. It will also supply visual art for AI‑generated audio stories in Copilot’s story mode. The rollout follows earlier releases of MAI-Voice-1 and MAI-1-preview, signaling Microsoft’s broader push to build its own AI stack while still offering OpenAI and Anthropic models for other services. Read more →

Microsoft has introduced MAI-Image-1, its latest text‑to‑image AI model that creates photorealistic images from user prompts. Though not yet released commercially, the model is available for testing on the community platform LMArena, where it currently ranks among the top ten AI image generators. Microsoft says the tool is engineered to avoid repetitive or generic outputs and to deliver fast, high‑quality results, with particular strength in realistic lighting and landscape rendering. The company plans to integrate MAI-Image-1 into its broader Copilot and Bing Image Creator services after further development. Read more →

Microsoft announced the debut of MAI-Image-1, its first internally developed AI model for generating images. The new system, which emphasizes photorealistic output and natural lighting, is currently being tested on LMArena and is slated for integration with Copilot and Bing Image Creator. This rollout continues Microsoft’s strategy of expanding its own AI portfolio, which already includes MAI-Voice-1 and MAI-1-preview, and signals a reduced reliance on its longstanding partnership with OpenAI. Read more →

Microsoft AI announced its first internally developed text‑to‑image model, MAI‑Image‑1, describing it as a major step in its AI roadmap. The system is built to deliver photorealistic images quickly and was placed in the top 10 on the LMArena benchmark after feedback from creative professionals. Microsoft highlighted safety guardrails and positioned the new model alongside its other AI offerings, including MAI‑Voice‑1 and MAI‑1‑preview, while noting ongoing collaborations with Anthropic and a complex relationship with OpenAI. Read more →

Microsoft has introduced its own AI models, MAI‑1‑preview and MAI‑Voice‑1, aiming to provide specialized, consumer‑focused capabilities within its Copilot suite. The move, highlighted by AI lead Mustafa Suleyman in a discussion with The Verge, signals a strategic shift toward greater independence from OpenAI and a focus on models that excel at everyday queries and instruction following. Early deployments include Copilot Daily, Podcasts, and a public testing interface in Copilot Labs, with plans to expand text use cases in the coming weeks. Read more →

Microsoft’s AI division announced its first homegrown AI models: MAI-Voice-1, a speech model that can generate a minute of audio in under one second on a single GPU, and MAI-1-preview, a text model trained on roughly 15,000 Nvidia H100 GPUs. The new models already power features like Copilot Daily and are being tested in Copilot Labs and on the LMArena benchmarking platform. Microsoft says the models are designed for consumer‑focused experiences, offering fast, high‑quality audio generation and instruction‑following text responses, and will be rolled out in select Copilot scenarios. Read more →

Microsoft has unveiled two new artificial‑intelligence models that were built entirely by its own teams. The speech‑generation model, MAI-Voice-1, powers features such as Copilot Daily and Podcast, while the text‑focused MAI-1-preview is the company’s first end‑to‑end foundation model. Both were designed for efficiency, with MAI-Voice-1 running on a single GPU and MAI-1-preview trained on roughly 15,000 Nvidia H‑100 GPUs. Microsoft plans to open MAI-1-preview for public testing and to integrate it into select Copilot scenarios, signaling a strategic move toward greater independence from external AI providers. Read more →