GPT-5.2 Codex
#2Terminal Bench 2.0
by OpenAI
+
+
+
+
About
GPT-5.2 Codex is the coding-specialized variant of the GPT-5.2 family, built for software engineering workflows including automated code generation, refactoring, and agentic coding tasks. It extends GPT-5.2's improvements in general intelligence and tool-calling into code-specific use cases.
+
+
+
+
Timeline
ReleasedJan 14, 2026
+
+
+
+
License & Family
License
Proprietary
Performance Overview
Performance metrics and category breakdown
Overall Performance
2 benchmarks
Average Score
55.8%
Best Score
66.5%
High Performers (80%+)
0Top Categories
Coding
55.8%
+
+
+
+
All Benchmark Results for GPT-5.2 Codex
Complete list of benchmark scores with detailed information
| Terminal Bench 2.0 | Coding | 66.50 | 66.5% | Unverified | |
| SWE-rebench | Coding | 45.00 | 45.0% | Unverified |