The AI image wars are entering a new phase. For the past year, most comparisons have focused on text rendering, typography, and prompt adherence. Those tests are important, but they're also becoming less useful as every major model gets better at spelling words and placing labels on a page. So we decided to test something different. We took the latest image models from OpenAI, Google, Microsoft, and Ideogram and ran them through four challenges designed to measure very different capabilities:...
Microsoft Build 2026 recap, from Windows to Copilot, all AI