OpenAI GPT-5.2

GPT-5.2 introduces a three-tier intelligence architecture—Instant, Thinking, and Pro—that automatically calibrates response complexity to the task at hand, balancing speed and reasoning depth while optimizing compute costs. The model is engineered for complex, multi-step professional workflows, showing marked improvements in creating spreadsheets, building presentations, writing code, perceiving images, understanding long contexts, and using tools in extended projects. New capabilities like response compaction enable effective handling beyond its base context window, while xhigh reasoning effort delivers maximum analytical depth for critical decisions.

Released in December 2025, GPT-5.2 achieves 70.9% expert-level performance on the GDPval benchmark, surpassing 38% fewer errors compared to its predecessor and outperforming competitors in key knowledge work categories. It crosses the 90% threshold on ARC-AGI-1 for genuine reasoning ability and reaches 100% on AIME 2025 competition math and 92.4% on GPQA Diamond science questions. A specialized Codex variant further optimizes the model for agentic coding and defensive cybersecurity applications. Cached input discounts enable significant cost savings on repetitive workflows, making the model practical for sustained enterprise use.