GPT-5 Codex is a coding-focused variant of GPT-5 built for interactive development and long-running, autonomous engineering work. It excels at feature implementation, debugging, large-scale refactors, and code review, with higher steerability and tighter adherence to developer instructions for cleaner, production-ready code.
Context Window
256.0K
Max Output
32.8K
Input Price (Auto)
$1.25/1M
Output Price (Auto)
$10.00/1M
Cache Read (Auto)
$0.13/1M
Performance metrics and benchmarks
Sourced from Artificial Analysis.
Intelligence Index
21.8
Auto routing is available for this model. Explicit provider selection is not available.
Loading provider options…
Coding Index
21.2
Agentic Index
52.7
GPQA Diamond
Graduate-level scientific reasoning
68.6%
Better than 59% of models compared
HLE
Humanity's Last Exam
5.8%
Better than 53% of models compared
IFBench
Instruction-following benchmark
45.0%
Better than 60% of models compared
T²-Bench Telecom
Conversational AI agents in dual-control scenarios
0.0%
Better than 3% of models compared
AA-LCR
Long context reasoning evaluation
63.7%
Better than 86% of models compared
GDPval-AA
Economically valuable tasks
35.7%
Better than 95% of models compared
CritPt
Research-level physics reasoning
5.1%
Better than 95% of models compared
SciCode
Python programming for scientific computing
37.8%
Better than 76% of models compared
Terminal-Bench Hard
Agentic coding and terminal use
12.9%
Better than 55% of models compared
AIME 2025
American Invitational Mathematics Examination 2025
48.3%
Better than 47% of models compared
MMLU-Pro
Professional and academic subject knowledge
82.0%
Better than 78% of models compared
AA-Omniscience Accuracy
Proportion of correctly answered questions
38.7%
Better than 98% of models compared
Last updated May 15, 2026
Artificial AnalysisLiveCodeBench
Contamination-free coding benchmark
54.3%
Better than 61% of models compared
AA-Omniscience Hallucination Rate
Rate of incorrect answers among non-correct responses
74.1%
Better than 73% of models compared
