GPT-5.2-Codex
GPT-5.2-Codex adds stronger long-horizon coding, Windows performance and cyber capability, with invite-only access to more permissive cyber models for vetted defenders.
GPT-5.2 tuned for agentic coding with context compaction and large refactors; Terminal-Bench 2.0 64.0% vs 62.2% for GPT-5.2. First OpenAI release with a trusted-access preview for cybersecurity professionals. API on 2026-01-14.
- Date
- Thursday, 18 December 2025
- Lab
- OpenAI
- Kind
- model
- Access
- closed API
- Price
- API from 2026-01-14
Figures
| Measure | Value | Measured by |
|---|---|---|
| Terminal-Bench 2.0 | 64.0% GPT-5.2 62.2% | company |
Codex usage reportedly doubled after its launch; the Codex app launch post cites over a million developers in the prior month (2026-02-02).
Sources
- openai.com/index/introducing-gpt-5-2-codex/
- simonwillison.net/2025/Dec/19/introducing-gpt-52-codex/
- deploymentsafety.openai.com/gpt-5-2-codex/
- developers.openai.com/api/docs/changelog
This record was checked against its sources on 6 October 2026. How we check