OpenAI Rolls Out GPT-5.4 Across ChatGPT, Codex, and the API

OpenAI has launched GPT-5.4, its newest model family update, and is now rolling it out across ChatGPT, Codex, and the API. The release lands as the company continues to refine how its flagship models handle coding, reasoning, and workplace tasks.

• GPT-5.4 is available across ChatGPT, Codex, and the API.
• OpenAI says the model improves reasoning, coding, and professional work outputs.
• Paid ChatGPT users get access to GPT-5.4 Thinking, with legacy access to GPT-5.2 Thinking temporarily preserved.

A Fresh Push for Professional AI

OpenAI says GPT-5.4 is designed to deliver better results in tasks that matter to developers and knowledge workers, including coding, document generation, and agent-style workflows. In its launch material, the company describes the model as a step up in general reasoning and coding ability, with the rollout extending across multiple products instead of staying limited to one surface.

That matters because GPT releases are no longer just about chatbot behavior. They now shape how businesses use AI inside coding tools, agent systems, and enterprise workflows. GPT-5.4’s launch suggests OpenAI wants the latest model to be a broad platform layer, not just a consumer-facing upgrade.

What Changes for ChatGPT and Codex

For ChatGPT, GPT-5.4 Thinking is being made available to Plus, Team, and Pro users, while Enterprise and Edu customers can enable early access through admin settings. OpenAI also said GPT-5.2 Thinking will remain available for three months in the legacy model picker before being retired on June 5, 2026.

In Codex, the company says GPT-5.4 includes experimental support for a 1M context window, a notable shift for longer coding and agentic tasks. OpenAI is also positioning the model as part of a broader simplification of model choice inside its developer tools.

Pricing and Benchmarks Signal a Higher-End Model

OpenAI is charging more per token for GPT-5.4 in the API than GPT-5.2, while saying improved token efficiency may reduce the total amount of usage needed for some tasks. The company also introduced GPT-5.4 Pro for developers who need more capability on complex jobs.

OpenAI’s own benchmark table shows GPT-5.4 ahead of earlier models on several professional, coding, tool-use, and reasoning tests, though benchmark gains do not always translate cleanly into real-world performance. Still, the company is clearly framing GPT-5.4 as a premium step forward rather than a routine maintenance release.

Why This Release Matters

The broader significance of GPT-5.4 is strategic. OpenAI is showing that its newest model releases are increasingly tuned for workplace productivity and software development, not just conversational polish. That puts more pressure on rivals competing for enterprise budgets and developer mindshare.

It also suggests OpenAI is continuing to separate “Thinking” and “Instant” model lines, which may become a defining feature of how users navigate ChatGPT going forward.

What to Watch

The key question now is whether GPT-5.4’s stronger benchmark results translate into noticeably better day-to-day performance for paying users and developers. It will also be worth watching how quickly OpenAI expands access beyond the initial rollout, and whether the model’s higher pricing changes adoption in the API.

Another point to monitor is whether GPT-5.4 becomes the new default across more OpenAI products, and how long the company keeps legacy models available as it pushes users toward the newer line.