GPT-5.2

GPT-5.2 shipped on 11 December 2025, four weeks after GPT-5.1, as OpenAI's frontier model for complex professional work. It keeps the 400,000-token context window and 128,000-token output ceiling of the generation, takes text and images and returns text, and moves the knowledge cutoff forward eleven months to 31 August 2025. Its distinguishing feature is a fifth reasoning setting. Where GPT-5.1 offered none, low, medium and high, GPT-5.2 added "xhigh" — a tier above what the family had allowed until then. The price rose accordingly: $1.75 per million input tokens and $14 per million output, 40% above GPT-5.1 on input and output alike, with cached input at a tenth of the input rate. One detail is worth recording because it is rare enough that vendors seldom state it plainly. On 21 January 2026 OpenAI rebuilt its inference stack and reported that GPT-5.2 and GPT-5.2-Codex now ran roughly 40% faster, explicitly noting that the model and its weights were unchanged. The same speed-up therefore reached every developer without a migration, a re-evaluation or a new snapshot. The Codex variant was shut down on 23 July 2026 and gpt-5.2-chat-latest on 10 August 2026; the base model has no announced shutdown date.

▸Videos

No videos yet.

▸Reviews