GPT-5.4

OpenAI's most capable and efficient frontier model for professional work, released March 2026. GPT-5.4 is the first mainline reasoning model to incorporate the frontier coding capabilities of GPT-5.3-Codex while adding native, state-of-the-art computer-use capabilities. It supports a 1M+ token context window (922K input, 128K output) and configurable reasoning effort (none, low, medium, high, xhigh). It scored a record 83% on GDPval for knowledge work, 57.7% on SWE-Bench Pro (surpassing GPT-5.3-Codex), and 75% on OSWorld-Verified (exceeding the 72.4% human baseline). Features include tool search for efficient large tool surfaces, native computer use for operating software through UI, and the full modern agent stack. GPT-5.4 is 33% less likely to make individual claim errors and 18% less likely to have response-level errors vs GPT-5.2.