GPT-5.4 is OpenAI's flagship reasoning model in ChatGPT, featuring native computer use for controlling software via screenshots, a 1-million token context window, and superior performance in professional tasks like spreadsheet modeling, coding, presentation creation, and research. It excels in OSWorld-Verified benchmarks at 75%, outperforms prior models in factual accuracy with 33% fewer hallucinations, supports tool use including web search and data analysis, and allows interruption during thinking for refined outputs.[1][2]