5.4 sooner than you Think.
— OpenAI (@OpenAI) March 3, 2026
Update 06/03/26 – 09:09 am (IST): OpenAI has now officially launched GPT-5.4, with availability across ChatGPT, the API, and Codex, while GPT-5.4 Pro is also going live for users who need extra performance on more complex tasks. The company is pitching it as its most capable and efficient model for professional work yet (duh!), with upgrades to reasoning, coding, tool use, and native computer control.
Update 05/03/26 – 11:45 am (IST): Fresh chatter on X suggests GPT-5.4 is now being tested on Arena, though the exact variant is still unclear.
The build is being spotted under the “Galapagos” codename in Arena battles, so more details should surface as more users get access.
Meanwhile, others predict it’ll officially drop tomorrow.
Original article published on March 4, 2026, follows:
OpenAI rolled out a meaningful tweak to ChatGPT with the GPT-5.3 Instant update, and barely an hour after the announcement, they teased the next big upgrade in line. The company posted a simple line on X yesterday: “5.4 sooner than you think.”
The GPT-5.3 Instant version went live on March 3, bringing noticeable improvements to the everyday conversations most users have with ChatGPT. OpenAI designed this update to address long-standing complaints about the model’s tone and behavior. Responses should now feel more natural and direct, with far fewer of those overly cautious preambles or moralizing warnings that used to break the flow.
The company claims hallucinations have dropped too, with a 26.8% reduction when the model pulls in web search results, and even better accuracy on pure knowledge questions. Web integration works more smoothly as well; instead of just dumping lists of links, the model weaves relevant info into clearer, more contextual answers. Writing tasks benefit from the changes, producing smoother prose whether you’re drafting a quick email or working on something more creative.
This update is already available to everyone using ChatGPT, and developers can access it through the API under the name gpt-5.3-chat-latest. The previous GPT-5.2 Instant will stick around as a legacy option for paid users until June 3, 2026, giving time for a smooth transition. OpenAI has said they’ll bring similar improvements to the Thinking and Pro modes in the coming weeks.
But let’s come back to the bigger update that’s around the corner. The quick tease about GPT-5.4 fits a pattern we’ve seen building over the past week or so. Back in late February, references to GPT-5.4 started slipping into Codex pull requests and even the model selector.
Even earlier, users reported getting pulled into A/B tests for what seemed like GPT-5.3, with prompts hinting at a new version in the app. Now, with GPT-5.3 Instant out the door and the immediate nod to 5.4, OpenAI looks to be accelerating its iteration cycle.
This push comes amid stiff competition. Google recently unveiled its own updates, including cost-efficient models like Gemini 3.1 Flash-Lite, keeping the pressure on for better daily usability. OpenAI seems to be prioritizing fixes that matter to regular users — less “cringe,” more straight answers — over just raw benchmark chasing.
Details on what GPT-5.4 will actually deliver remain scarce, but one widely shared post by entrepreneur Alex Finn claimed GPT-5.4 could introduce a context window of up to two million tokens and potentially include open-source elements. Those claims remain unverified, so take the details with a grain of salt.
Whatever it brings, the timing suggests OpenAI isn’t planning to slow down. OpenAI hasn’t shared any official timeline, but keep an eye out for the new model in the coming days, possibly even tomorrow. At least that’s what some users are speculating, considering the capital ‘T’ in OpenAI’s post, which some believe might point to ‘Today,’ or ‘Thursday’.
For the moment, though, you can give the new GPT-5.3 Instant model a spin and see if you notice any improvements over the previous version.
Featured image generated with AI


