GPT-5.4 expands OpenAI’s vision for autonomous AI agents

OpenAI is launching GPT-5.4, a more capable AI model that combines reasoning and coding with native computer-use capabilities, enabling it to operate a user’s computer across apps. It includes GPT-5.4 Thinking for ChatGPT to outline complex tasks and let users tweak results mid‑response, and it improves web-browsing, tool and API usage, and multi-source research. OpenAI says GPT-5.4 is more factual (claims 33% less likely to be false than GPT-5.2) and is rolling out via API, Codex, and the ChatGPT platform, with specialized Pro variants for higher performance and enterprise use as the company moves toward a future where AI agents can autonomously complete complex tasks online and in software.