OpenAI Launches GPT-5.4 with Pro and Thinking Versions

OpenAI released GPT-5.4 across ChatGPT, the API, and Codex. The model's ARC-AGI-2 benchmark score jumped from 52.9% to 73.3%, a test specifically designed to resist surface-level pattern matching. GPT-5.3-Codex had launched earlier as OpenAI's most capable agentic coding model, combining Codex and GPT-5 training stacks. OpenAI's product organisation was renamed 'AGI Deployment,' the first time AGI appeared as a formal category in its org chart.