Codex CLI vs Claude Code

Feature	Codex CLI	Claude Code
Tagline	Lightweight coding agent that runs in your terminal	An agentic coding tool that lives in your terminal
Website	https://openai.com/codex	https://claude.ai/code
Language	rust	typescript
Release Date	April 30, 2025	February 22, 2025
Last Updated	June 29, 2026	September 4, 2026
Provider Type	AI Lab	AI Lab
MCP Support	Yes	Yes
Multi-Agent Support	Yes	Yes
Git Integration	deep	deep
Multi-File Editing	Yes	Yes
Auto Test Running	Yes	Yes
Sandbox / Permissions	Yes	Yes
Image / Multimodal Input	Yes	Yes
Voice Mode	Yes	Yes
IDE Integration	VS Code, Zed (via ACP)	VS Code, JetBrains
Hooks / Extensions	Yes	Yes
Web Search	Yes	Yes
Local Model Support	Yes	Yes
Context Window	Model-dependent	1M tokens
Terminal Bench	#1 · 58.2% ±2.8 (Terminal-Bench 4.0) 1	#2 · 57.9% ±3.8 (Terminal-Bench 4.0) 2
AA Coding Agent Index	67% · GPT-6 Astra (max)	70% · Fable 5.1 (max, with fallback)
Supported Models	All frontier GPT, Codex models: GPT-5.6 Sol / Terra / Luna, GPT-5.5, GPT-5.4 pro, GPT-5.3-Codex, optional custom models.	Claude Fable 5 / Mythos 5 (Restricted Availability), Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5
Supported Subscriptions	ChatGPT	Claude Code
Pricing Model	Subscription	Subscription
Free Tier	No	No
Price Range	$20-200/mo	$17-200/mo
Source Code Availability	Open Source (Apache-2.0)	Proprietary Software
Source Code	https://github.com/openai/codex	https://github.com/anthropics/claude-code
Development State	active	active

¹ Codex CLI tested with GPT-6 Astra on Terminal-Bench 4.0.

² Claude Code tested with Fable 5.1 on Terminal-Bench 4.0.