ChatGPT has rolled out GPT-5.4 mini for free users: all the details

This was announced by OpenAI.

OpenAI has unveiled two new compact models – GPT-5.4 mini and GPT-5.4 nano. The company stated that these are currently the most powerful small models in its range, with the main focus being on GPT-5.4 mini, which has already been added to ChatGPT, Codex and the API.

For Free and Go users, the GPT-5.4 mini model is available via the “Thinking” feature in the “+” menu. For other ChatGPT users, it is used as a fallback option once the GPT-5.4 Thinking limit has been reached. Separately, OpenAI explains that at the free tier, once the GPT-5.3 limit is exhausted, chats automatically switch to the mini version of the model until the limit is reset.

The company emphasises that GPT-5.4 mini is faster than the previous GPT-5 mini and performs better on programming tasks, working with tools, and multimodal understanding. According to OpenAI, the model demonstrates significant progress in coding, rapid editing, navigating large codebases, interface generation and debugging.

Among the technical capabilities of GPT-5.4 mini, OpenAI highlights support for text and image inputs, function calls, web search, file search, computer usage and other tools. The model’s context window is 400,000 tokens.

The company also highlights the benchmark results. According to official data, GPT-5.4 mini has come close to the full-scale GPT-5.4 in a number of specialised tests, notably OSWorld-Verified and SWE-Bench Pro, whilst significantly outperforming GPT-5 mini.

OpenAI has made GPT-5.4 mini available to developers via API at a price of $0.75 per 1 million input tokens and $4.50 per 1 million output tokens. In Codex, this model uses only 30% of the GPT-5.4 quota, allowing it to perform simpler tasks at roughly a third of the cost of the full-sized model.

Alongside the mini, the company has also announced GPT-5.4 nano – the smallest and cheapest model in the new range. It is not available directly in ChatGPT and is intended solely for use via the API. It costs $0.20 per million input tokens and $1.25 per million output tokens. OpenAI recommends the nano for data classification, information retrieval and support agents with simpler tasks.

In this way, OpenAI is expanding access to the new model not only for paying users but also for a wider audience without a subscription. For ChatGPT, this means that some of the features previously associated mainly with premium access are now being rolled out in a simplified format to the free tier. This could intensify competition between AI services, particularly in the mass user segment. This conclusion follows from the new access scheme described by OpenAI.