
Chinese Image Models Are Going Global, and This Month the Conversation Changed
Chinese image models are going global. How the benchmarks, the comment sections, and Washington changed this month.
Category
145 articles

Chinese image models are going global. How the benchmarks, the comment sections, and Washington changed this month.

Prediction markets are betting on who wins image AI. What the lopsided odds actually measure, and how to read them.

Why humans and detectors are losing the AI image arms race, and what provenance and community actually do.

Qwen Image 2.1, a 7B open-weights model, rivals Nano Banana 2.0 and runs on a laptop. What that means for image generation.

About a cent per image, aimed at advertisers: what Meta Muse's pricing means for ad creative and the two-lane image market.

OpenAI retired the DALL-E GPT on August 30, 2026. What changed, why now, and what the name meant for AI art.

Object isolation, in-image text translation and Workspace integration: what Google Pics means for Canva and Adobe Express.

Image 2.0 ranks second on both image arenas with region editing, 5-image fusion and smart resize. The editing-first bet, explained.

Images 2.5 cuts latency up to 50%, adds Sketch, and stabilizes multi-turn edits. What changed and who should care.

SenseTime's 8B model, Ant's 6B model, and Shusheng's InternLumina-U2 were open-sourced in quick succession. This article breaks down the differences among the three approaches and the real impact of the open-source wave on creators.

HyImage 3.5 preview improves by about 30% in blind tests, supports 5 reference images and 2K output, has been integrated into WorkRally and used in production on Xingbiao. This article breaks down its capabilities and usage paths.

GPT Image 2.5, Midjourney V8.2, Nano Banana 2, Flux 2 and more, ranked with prices and picks by job.