| Claude Code | Codex CLI | |
|---|---|---|
| arenaElo | 1420 | 1408 |
| humanEval | 93.1 | 95.5 |
| sweBench | 64.3 | 70 |
| mmlu | 92.3 | 88.6 |
| gpqa | 65.2 | 60.8 |
| aime | 88 | 86 |
| Price in/out ($/M) | $15 / $75 | $9 / $36 |
Claude Code rides on Opus 4.7: stronger long-horizon agents, 1M context, best-in-class MCP ecosystem. Codex CLI rides on GPT-5.4 Codex: highest HumanEval, tight OpenAI tool-calling, faster loops. Pick Claude Code for multi-hour refactors; pick Codex CLI for fast PR factories.