Claude Code scores higher. Codex costs ~7× less per task

I ship my own code with Claude Code. The question I keep getting from other freelancers isn't "which model is smarter?" — it's "which agent seat burns less money for the same week of client work?"

Artificial Analysis's Coding Agent Index v2.0 (FrontierCode + Terminal-Bench + SWE-Atlas-QnA) puts Claude Code with Claude Opus 5.5 (max) at 62, vs Codex with GPT-6 Astra (max) at 55. Same board, different bill: that Claude Code run averages $11.42 API cost and ~1 hour wall time per task; Codex Astra averages $7.04 and 24.8 minutes. Codex with GPT-6.1 Sol sits near 54 for about $1.50 per task. Those are pay-per-token API costs, not your Pro/Max or ChatGPT plan price: https://artificialanalysis.ai/agents/coding-agents/comparisons/claude-code-vs-codex

How I actually use that:

1. Hard multi-file product work → Claude Code (the seat I pay for and supervise).

2. Fast, cheaper passes and second opinions → a Codex-class seat when the quota or bill says so.

3. Research, planning, and QA → separate bots. They don't own the repo.

Benchmarks don't ship clients. Your monthly ceiling does.

When a coding agent burns tokens, do you optimize for score, wall time, or monthly bill first?

More at huzaifatahir.dev