OpenAI is preparing to launch the next version of its large language model (LLM), GPT-5.4. Although no exact launch date has been revealed yet, the ChatGPT-maker says it’s coming “sooner than you think.” While you wait, you get the option of GPT-5.3 Instant, which is specifically designed to address some of the well-documented shortcomings of previous GPT-5 iterations around its tone and accuracy. Meanwhile, online reports suggest that GPT-5.4 will offer a larger context window and extreme reasoning capabilities.

According to the information, GPT-5.4 will unlock a larger context window for users. More precisely, it is reported that it will come with one million tokens (some reports suggest that 2 million tokens are also coming) which is a huge jump from the 400,000 tokens available in GPT-5.2. In other words, it will be able to capture more data in a single prompt thereby speeding up coding tasks as well as in-depth research.

In fact the new model appears to be designed for research, with the report hinting at an “extreme” reasoning mode that would allow ChatGPT to answer more complex questions, although this is likely to come at the expense of time and computation. More information is awaited.

Meanwhile, OpenAI has released GPT-5.3 Instant. The announcement comes at a time when the ChatGate-maker is facing criticism for signing a deal with the US military, while rival Anthropic has decided to pull out of it on moral and ethical grounds. The agreement makes OpenAI the default choice for AI deployment in Pentagon operations, although the company has paused deployments at the NSA and other defense agencies for a while so it can make some changes before moving forward on the contract. The crux of the controversy is the use of AI for mass surveillance and the development of autonomous weapons.

The loss of Anthropic did not bring any short-term benefits to OpenAI. Ever since Sam Altman swooped in to sign the same contract that Anthropic rejected, hoping it would ease tensions between the US government and the AI ​​industry, ChatGPT users have been flocking to the cloud and other AI chatbots, resulting in a nearly 300 percent uninstall over the weekend (even as the cloud became the number one app on the Apple US App Store).

See also

Tech Hub

Apple may bring its own in-house cellular modem by 2025, but it still has a long way to go

While Sam Altman has acknowledged that he may have handled the situation hastily and carelessly, and while signing the agreement may have seemed opportunistic to outsiders, the comments have done little to placate users and even some OpenAI employees, who have either quit or shown support for Anthropic, protesting the use of AI in military surveillance.

OpenAI will be hoping that in the interim GPT-5.4 and GPT-5.3 instantiations (which are supposed to be less troublesome and more useful) will convince existing users to stay and those who left to come back.