OpenAI is preparing to release GPT-5.4, the next version of its large language model powering ChatGPT. While an exact launch date has not been announced, the company says the update is coming “sooner than you think.”
GPT-5.4: Key expected upgrades
- Much larger context window:Reports suggest GPT-5.4 could support around 1 million tokens, with some estimates going as high as 2 million tokens.
The previous version, GPT-5.2, supports about 400,000 tokens.
A larger context window means the model can process much more text, code, or documents in a single prompt.
- Extreme reasoning mode:The model is expected to include a new reasoning capability designed to tackle complex research problems and advanced queries.
This mode may take more time and computing power but should produce deeper, more accurate answers.
- Better for coding and research:With a larger context window and stronger reasoning, GPT-5.4 is expected to handle large codebases, research papers, and long conversations more efficiently.
GPT-5.3 Instant released meanwhile
While users wait for GPT-5.4, OpenAI has introduced GPT-5.3 Instant. This version is meant to fix some criticisms of earlier GPT-5 models, particularly:
overly robotic or awkward tone
occasional accuracy issues
need for faster responses
The “Instant” model focuses on being quicker and more practical for everyday tasks.
Context: controversy around OpenAI
The release announcements come during criticism of Sam Altman and the company after it signed a deal with the US Department of Defense for AI deployment. Rival AI company Anthropic declined a similar contract due to concerns about potential uses such as mass surveillance or autonomous weapons.
Following the deal:
some users reportedly uninstalled ChatGPT
many switched to Anthropic’s Claude chatbot
some OpenAI employees criticised or left the company
Altman later acknowledged the deal may have appeared rushed and opportunistic.
OpenAI is expected to rely on GPT-5.4 and GPT-5.3 Instant to rebuild user confidence and bring users back to ChatGPT.
