Additionally, there are improvements to visual understanding; it can now more carefully analyze images up to 10.24 million pixels, or up to a 6,000-pixel maximum dimension. OpenAI also claims responses from this model are 18 percent less likely to contain factual errors than before.

ChatGPT reportedly lost some users to competitor Anthropic in recent days, after OpenAI announced a deal with the Pentagon in the wake of a public feud between the Trump administration and Anthropic over limitations Anthropic wanted to impose on military applications of its models. However, it’s unclear just how many folks jumped ship or whether that led to a substantial dip in the product’s massive base of over 900 million users.

To take advantage of the situation, Anthropic rolled out the once-subscriber-only memory feature to free users and introduced a tool for importing memory from elsewhere. Anthropic says March 2 was its largest single day ever for new sign-ups.

OpenAI needs to compete in both capability and cost and token efficiency to maintain its relative popularity with users, and this update aims to support that objective.

GPT-5.4 is available to users of the ChatGPT web and native apps, Codex, and the API starting today. Subscribers to Plus, Team, and Pro are also getting GPT-5.4 Thinking, and GPT-5.4 Pro is hitting the API, Edu, and Enterprise.