{
  "id": 11217664,
  "title": "OpenAI Codex vs Claude Code: The Better AI Coding Agent Depends on More Than Benchmarks",
  "url": "https://urgent.news/2026/10/01/openai-codex-vs-claude-code-the-better-ai-coding-agent-depends-on",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-01T07:54:22.000Z",
  "source": {
    "name": "TechRepublic",
    "slug": "techrepublic",
    "url": "https://www.techrepublic.com/article/news-codex-vs-claude-code/"
  },
  "original_language": "en",
  "account": "OpenAI Codex and Anthropic Claude Code, two leading AI coding agents, recently competed in coding-agent benchmarks, showcasing their performance, cost, speed, and deployment controls. While Anthropic's Claude Code currently leads with a higher composite score, OpenAI's Codex holds its own at a fraction of the cost. Recent benchmark scores indicate Claude Code scored 68 on Artificial Analysis' Coding Agent Index, using Sonnet 5.5 at matched xhigh effort, while Codex scored 63 with GPT-6.1 Sol. However, the gap narrows at matched xhigh effort, where both agents score 63. Notably, Codex demonstrates faster task completion and lower API cost. Anthropic introduced Sonnet 5.5 on September 28, which reportedly improves throughput by over 30% and reduces task costs by up to 30% for most workloads. OpenAI followed suit on September 29 with GPT-6.1 Sol, designed for complex coding tasks at lower cost than its previous GPT-6 Astra. These pricing and performance shifts emphasize the importance of evaluating cost per completed task alongside token prices. Both agents underwent rigorous testing for security, with Claude Code users approving 93% of permission prompts, while Anthropic reported a 17% false-negative rate in cutting \"overeager\" actions. As teams consider adopting either agent, they should thoroughly test both platforms against their repository targets, verify credential scope, and evaluate command-approval policies before granting broader access.",
  "summary": "Codex and Claude Code split the latest coding-agent benchmarks on performance, cost, speed, and deployment controls. The post OpenAI Codex vs Claude Code: The Better AI Coding Agent Depends on More Than Benchmarks appeared first on TechRepublic .",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}