Overview
- The GPT-5.2 family is available in three modes—Instant for speed, Thinking for deeper reasoning, and Pro for research‑grade output—rolling out first to Plus, Pro, Business, and Enterprise tiers, with API access live.
- OpenAI reports benchmark gains including a 70.9% win/tie rate on GDPval knowledge‑work tasks and 55.6% on SWE‑Bench Pro, alongside roughly 30–38% fewer factual errors than GPT‑5.1.
- The models target enterprise workflows with stronger tool use, agentic execution, and long‑context handling, with coverage citing a much larger context window and better vision on charts, diagrams, and interfaces.
- Microsoft confirmed GPT-5.2 and GPT-5.2‑Chat are generally available in Microsoft Foundry for enterprise deployment, with published per‑token pricing.
- The launch follows an internal “code red” as competition from Google’s Gemini 3 intensifies, and OpenAI signals forthcoming age gating with an adult mode targeted for early 2026.