Microsoft
Research driven AI models for 3D generation and visual understanding
Microsoft Research develops generative models that span 3D asset creation, image understanding, and multimodal reasoning. Models like TRELLIS bring high quality 3D generation from text and image prompts, reflecting deep investments in spatial computing and visual AI research. As a Runware provider, Microsoft models are available through a single inference pipeline alongside other creators.
Models by Microsoft
MAI-Image-2.6-Flash is the speed-optimized variant of Microsoft's MAI-Image-2.6 family, retaining its generation and editing capabilities for latency-sensitive production. It supports text-to-image generation, precise multi-reference editing, web grounding, and dynamic aspect ratios, with strengths in targeted object edits, layout adaptation, text updates, artifact cleanup, and visual consistency across iterations. It is designed for rapid prototyping, personalization, automated pipelines, interactive applications, and high-throughput creative production.
MAI-Image-2.6 is Microsoft AI's highest-quality image generation and editing model, built for precise creative control and polished production output. It supports text-to-image generation, controllable edits with up to five reference images, web grounding for current real-world context, and dynamic aspect-ratio selection with output up to 1.5K. Its strengths include text rendering, portraits, photorealism, commercial design, and maintaining people, products, styles, and scenes across complex edits.
TRELLIS.2 is a 4-billion-parameter 3D generative model that transforms 2D images into fully textured 3D assets with complex topology, sharp features, and physically based rendering materials. It uses a novel field-free sparse voxel representation and scalable structured latent space to enable efficient, high-resolution 3D generation from single image input.