OpenAI releases GPT-5.4, the most factual and most efficient model

PANews March 6 Report: According to Jin10, OpenAI officially launched GPT-5.4, claiming it is the most fact-based and efficient model—fewer tokens, faster speed. In ChatGPT, GPT-5.4’s reasoning has improved in deep web research, with better context retention during extended thinking—users can now interrupt the model during responses, add instructions, or adjust the direction. The guiding feature was launched on Android and web this week, with the iOS version coming soon. OpenAI developers say GPT-5.4 has native computer usage capabilities. Codex and API support up to 1 million context tokens, enabling optimal proxy coding for complex tasks, scalable tool search within larger ecosystems, and more efficient reasoning for long, tool-intensive workflows.