TAU2
Measured May 29, 2026Source
Score
0.46
Models13 1 
GPT-5.2 (Non-reasoning)
OpenAIOpenAIUnited States
A high-performance variant of the GPT-5 series optimized for speed and cost-efficiency. It is designed for general-purpose tasks, offering strong multimodal and coding capabilities without the overhead of extended reasoning chains.
CodingFastCheapLong contextMultimodal
Input / 1M tokens
$1.75Output / 1M tokens
$14.00Output tokens/s
64.69First-token seconds
0.67sSupported plans
6Benchmark history
Evaluations
Terminalbench Hard
Measured May 29, 2026Source
Score
0.32
Lcr
Measured May 29, 2026Source
Score
0.38
Ifbench
Measured May 29, 2026Source
Score
0.47
Aime 25
Measured May 29, 2026Source
Score
0.51
Scicode
Measured May 29, 2026Source
Score
0.4
Livecodebench
Measured May 29, 2026Source
Score
0.67
Hle
Measured May 29, 2026Source
Score
0.07
Gpqa
Measured May 29, 2026Source
Score
0.71
Mmlu Pro
Measured May 29, 2026Source
Score
0.81
Artificial Analysis Math Index
Measured May 29, 2026Source
Score
51
Artificial Analysis Coding Index
Measured May 29, 2026Source
Score
34.7
Artificial Analysis Intelligence Index
Measured May 29, 2026Source
Score
33.6
Plan availability
Products and plans that support this model
Codex
Codex
OpenAI's Codex is an AI coding agent that helps developers build and ship software by handling tasks end-to-end, including code generation, refactoring, and review. It integrates with IDEs, CLI, and web, powered by ChatGPT for individual and team workflows.
User ratings
Loading ratings...
Discussion
Thinking... Make sure you are connected to GitHub server