GPT-5.4 and GPT-5.4-Pro: OpenAI Redefines AI Performance

OpenAI has just unveiled its latest artificial intelligence models, GPT-5.4 and GPT-5.4-Pro. These new versions are now integrated into ChatGPT and Codex CLI, providing users with even more advanced technology. The knowledge cutoff date for these models is set at August 31, 2025, meaning they are up to date with information until that date.

One of the most impressive features of these new models is their ability to handle a context window of 1 million tokens, allowing for broader and more complex data processing. However, this technological advancement comes at a cost. The prices for the GPT-5.4 models are slightly higher than those of the previous series, GPT-5.2, with a notable increase in rates for usage exceeding 272,000 tokens.

In terms of performance, GPT-5.4 outperforms its coding-specialized predecessor, GPT-5.3-Codex, across all relevant benchmarks. This raises questions about the future of the Codex line, which may be integrated into the main version of GPT-5.4.

OpenAI has also emphasized the enhancement of GPT-5.4's capabilities in creating and editing professional documents such as spreadsheets, presentations, and text documents. In an internal benchmark, GPT-5.4 achieved an average score of 87.3% for spreadsheet modeling tasks, significantly surpassing the 68.4% score of GPT-5.2. This improvement is particularly relevant in the context of commercial applications, an area where Claude, another player in the industry, has recently focused its efforts.

To illustrate the creativity of these new models, OpenAI shared examples of drawings generated by GPT-5.4, including a pelican on a bicycle. Another drawing, created by GPT-5.4 Pro, took 4 minutes and 45 seconds to complete, at a cost of $1.55.