OpenAI rolls out GPT-5.4 Thinking and Pro across ChatGPT, API, and Codex

On March 5, 2026, OpenAI said on X that GPT-5.4 Thinking and GPT-5.4 Pro were beginning a gradual rollout across ChatGPT, the API, and Codex. The linked product page described GPT-5.4 as OpenAI's most capable and efficient frontier model for professional work, positioning it as a single model family for interactive chat, developer integration, and coding workflows.

The announcement highlighted four capabilities that matter for production use. OpenAI pointed to state-of-the-art coding, computer use, tool search, and a 1M-token context window. In the same thread, the company also said GPT-5.4 is its most factual and efficient model, emphasizing faster speed and fewer tokens. That framing suggests OpenAI is not presenting GPT-5.4 as a narrow reasoning upgrade, but as a general-purpose model line meant to cover a wide span of professional tasks.

• Rollout surface: ChatGPT, the API, and Codex.
• OpenAI's positioning: a frontier model for professional work.
• Capabilities cited in the announcement: coding, computer use, tool search, and 1M-token context.

The practical significance is that the same release touches both end-user interfaces and developer infrastructure. That can shorten evaluation cycles for teams that first test a model in ChatGPT and then move to API deployment or Codex-driven development flows. It also signals that OpenAI is continuing to converge reasoning, tool use, and software engineering assistance into a unified flagship offering rather than splitting those functions across separate product-specific models.

Primary sources here are OpenAI's March 5, 2026 X posts and the linked official announcement page. OpenAI explicitly described availability as gradual, so the most careful reading is that the company was starting a rollout, not claiming instant, uniform access across every surface and tier on day one. The official product page linked from the thread is Introducing GPT-5.4.