OpenAI has indicated that a new version of its large language model, GPT-5.4, is in development following a post on social media platform X.
5.4 sooner than you Think.
— OpenAI (@OpenAI) March 3, 2026
The upcoming model is anticipated to feature an “extreme” reasoning mode, which would allow it to dedicate more computational resources to complex problems, potentially expanding its capabilities for analysis.
Internal leaks suggest GPT-5.4 may include a two-million-token context window, an increase from GPT-5.3’s 400,000-token limit. This extended context would allow for processing large documents or entire codebases in a single session.
Additionally, a new “detail: original” API parameter is expected to enable pixel-level analysis of images by bypassing standard compression, according to leaked code analyzed by NxCode.
OpenAI’s recent release schedule has accelerated, with GPT-5.3 Instant launching on March 3 following GPT-5.3-Codex on February 4. GPT-5 debuted in August 2025.
GPT-5.3 Instant focused on reducing model refusals and improving conversational tone, achieving a reduction in hallucination rates by up to 26.8 percent compared to prior models, OpenAI stated in a blog post.
The two-million-token context figure remains uncorroborated by OpenAI’s own code, according to AwesomeAgents. Prediction markets currently assess a 55% probability of GPT-5.4 releasing by April 2026.
