AI RACE— The AI Race
AI model

GPT-5 Codex

OpenAI

Try it now ↗
Released09/2025
Speed0 tokens/s
Time to first token0.00s
Input price$1.25/1M
Output price$10/1M

Rankings

#103
AIM 41.5Peak #49
#121
AIM 80.1Peak #77
#3
AIM 92.2Peak #3

Review

GPT-5 Codex is a strong choice for engineers and scientists needing high precision in mathematics and coding, but is weak in open-ended dialogue and conversational AI. Suited for STEM professionals rather than general chat applications.

Strengths

  • Excels in mathematics and programming, ranking in the top 3 within its specialized domain
  • Decent logical reasoning and AI Agent handling, suitable for complex problems
  • Well-suited for STEM tasks requiring high precision

Weaknesses

  • Weak general chat (ranked #59), making it a poor choice for open-ended conversation or customer support
  • Lower AI Agent ranking compared to mathematics, limiting its ability to handle complex automation tasks
  • No clear standout strengths outside mathematics and programming

Use cases

Solving mathematical problems, data science, and analyticsProgramming and software development, particularly code generationLogical reasoning and solving structured, complex problems

Guides & videos

Access ChatGPT (chatgpt.com) with a ChatGPT Plus subscription or higher, then select GPT-5.3-Codex or GPT-5.5-Codex from the model menu. Use the model to write code, debug, and solve complex programming challenges. Describe requirements in natural language, and the model will generate code or explanations. Pro tip: include existing code context to help the model better understand the codebase; use the Codex CLI or IDE extensions (VS Code, JetBrains) to integrate it into your development workflow.

Reviews