GPT-4.1 is anticipated to take centre stage as OpenAI gets ready to unveil a new generation of AI models and features. Sources acquainted with the company’s plans claim that the next release will have what is probably going to be referred to as GPT-4.1, which is an improved version of the GPT-4o multimodal model.

GPT-4o, OpenAI’s flagship model with real-time reasoning capabilities spanning text, vision, and audio, was first released last year. GPT-4.1 will now build upon those foundations. As early as next week, OpenAI is anticipated to release scaled-down versions known as GPT-4.1 mini and nano in addition to the normal version.

But that’s not all. Along with an o4 tiny version that may be released even sooner, OpenAI is also developing the complete version of its o3 reasoning model. In a recent upgrade to the ChatGPT web interface, AI developer Tibor Blaho saw references to “o4 mini,” “o4 mini high,” and “o3” earlier today, suggesting that these models are almost ready for release. According to sources, the O3 and O4 mini are expected to debut next week, though dates are subject to change as usual.

Although it’s unclear whether this has anything to do with the future models, CEO Sam Altman hinted on X that an interesting feature would arrive today. OpenAI has recently delayed certain releases because of infrastructure constraints. By the time of publishing, the firm had not responded to a request for comment.

Customers “should expect new releases from OpenAI to be delayed, stuff to break, and for service to sometimes be slow as we deal with capacity challenges,” Altman previously warned users on X. Recently, OpenAI was forced to temporarily rate-limit user queries due to the popularity of its sophisticated image-generation capabilities. Altman went so far as to say, “our GPUs are melting,” alluding to the stress caused by the picture generator that is part of ChatGPT’s free plan.

The pace of OpenAI’s growth has not slowed in spite of these growing pains. The business is still pushing the limits of what generative AI can do, with GPT-4.1 and further models on the horizon.