Microsoft debuts MAI-Image-2-Efficient, a faster, cheaper variant of its flagship image model

What Microsoft announced

Microsoft has launched MAI-Image-2-Efficient, a lower-cost, higher-speed variant of its flagship text-to-image model, available immediately in Microsoft Foundry and the MAI Playground with no waitlist. It has been reported that the company is pricing the model at $5 per million text input tokens and $19.50 per million image output tokens — roughly a 41% cut versus MAI-Image-2’s image-output price — and that the new model runs about 22% faster and delivers 4x greater throughput efficiency per GPU on NVIDIA H100 hardware at 1024×1024. Microsoft says the model will also roll out across Copilot and Bing; it has been reported that the company claims MAI-Image-2-Efficient outpaces Google’s Gemini image variants on p50 latency by about 40%.

Who it’s for — and why this matters

This is a deliberate two-model play. MAI-Image-2-Efficient is aimed at high-volume, cost-sensitive production tasks — think product photos, marketing pipelines, UI mockups and real-time interactive tools — while MAI-Image-2 remains the precision instrument for top-tier photorealism, complex stylization and intricate in-image typography. It’s the same pricing strategy the industry has leaned on for text models, applied to images where pennies per output add up fast. Enterprise buyers, take note: cost-per-image can make or break a rollout. Who wouldn’t want to save half the bill on their image assembly line?

Fast cadence, big signals

Less than a month after MAI-Image-2’s debut, Microsoft shipping an optimized variant suggests the MAI Superintelligence team is operating like a fast-moving startup inside a trillion-dollar company. It has been reported that this speed — and the talk of an independent AI stack not reliant on OpenAI — is intentional. Can Microsoft really become self-sufficient in generative AI infrastructure? Maybe. The broader implication is clear: cheaper, faster image models lower the barrier to production-scale generative workflows, and the pressure on rivals to match price, latency and throughput just got dialed up.

Sources: venturebeat.com