Microsoft AI is kicking the value proposition up a notch. MAI-Image-2.6-Flash — a speed-focused variant of the flagship MAI-Image-2.6 — is now in public preview on Microsoft Foundry. It keeps the same core features: text-to-image generation, image editing, multi-image reference, live web retrieval, and automatic aspect ratio selection. The twist is that Microsoft is pushing latency and cost down hard.
Speed is the headline. Microsoft says Flash generates images 2.8x faster than GPT-Image-2 Medium and squeezes 72% more efficiency out of its GPUs. That doesn't mean you're just trading quality for quick turns, either. On Artificial Analysis's current image editing leaderboard, MAI-Image-2.6-Flash sits third with a 1311 Elo — just above GPT Image 2 high's 1309. For text-to-image, it's eighth at 1297 Elo. The non-Flash flagship MAI-Image-2.6 still holds the top spot in image editing and takes second in text-to-image.
Now the part that will get your attention: price. Microsoft Foundry lists Flash image output at $19 per million tokens, or exactly half the $38 per million for flagship 2.6. Artificial Analysis converts that to a per-image view: about $19.5 per 1,000 generated images, versus around $211 per 1,000 with GPT Image 2 high.