OpenAI and Google limit free Sora, Nano Banana Pro and Gemini 3 Pro use (www.techradar.com)

OpenAI and Google have quietly tightened free access to their newest generative models: OpenAI’s Sora 2 video generator is now capped at six video generations per day for free users (ChatGPT Plus and Pro limits remain unchanged, and extra generations can be purchased), while Google removed the previously advertised five daily prompts for Gemini 3 Pro free users and fixed Nano Banana Pro image generation to just two images per day. OpenAI’s Sora lead Bill Peebles cited “our GPUs are melting” to explain the rationing, and Google’s support notes say free “basic access” limits may change frequently; paid Google AI tiers (Pro: 100 prompts/day, Ultra: 500/day) are unaffected. The moves matter because they signal how rapidly rising demand and the high compute cost of image/video synthesis are forcing providers to ration resources and nudge users toward paid plans. Video and image generation is far more GPU‑intensive than text, so strict per‑user caps help protect capacity and keep services usable. For the AI/ML community, this is a practical reminder that model availability and experimentation will increasingly be shaped by infrastructure economics—expect more dynamic throttling, pricing changes, and tighter free tiers as companies balance load, cost, and product adoption.