ChatGPT briefly went offline on Sunday, hours after OpenAI CEO Sam Altman asked users to slow down image generation.
OpenAI had unlocked ChatGPT's native image generation capabilities, allowing the chatbot to create more nuanced and accurate images than when it relied on DALL-E to generate images.
Soon after, social media users began using ChatGPT to convert their real-life images in Studio Ghibli style.
This trend led to OpenAI limiting image generation capabilities even for ChatGPT Plus and Pro subscribers due to the growing load on the company's GPUs.
While the world was going gaga over ChatGPT's new image generation capabilities, Google was quietly rolling out its latest language model, Gemini 2.5 Pro, to all users.
This is a reasoning model similar to OpenAI's o3 Mini or DeepSeek R1, meaning it tries to mimic human-level reasoning before answering a query.
The new model is currently the top-ranked chatbot on LMArena, an open-source platform developed by researchers at UC Berkley SkyLab.
Gemini 2.5 Pro excels in reasoning, coding, math, and science, with a 1M-token context window (~750K words), surpassing Claude 3.7 Sonnet (500K tokens/~375K words) and o3 Mini (200K tokens/~150K words).
How to create ‘Ghiblified’ images using ChatGPT for free?
However, amid a "biblical demand" for Studio Ghibli-style images, the big question is whether Google's new language model can generate images inspired by the Japanese animation company.
Gemini 2.5 Pro's release blog does not clarify if it has native image generation or specify an external model for this task. It highlights the model's ability to process diverse inputs (text, audio, images, video, codebases) for complex problem-solving.
Gemini 2.5 Pro uses Google’s Imagen 3 for image generation, contrasting with GPT-4o’s superior native capabilities in ChatGPT.
Gemini 2.5 Pro cannot "Ghiblify" images, responding with an error: "Studio Ghibli style tool unavailable.