OpenAI Codex 是什麼?
OpenAI Codex is OpenAI's autonomous coding agent product — distinct from the original 2021 Codex model that powered early GitHub Copilot — capable of taking on a coding task described in natural language and carrying it out across a codebase, from planning and writing code through running tests. It's accessed through ChatGPT's Codex interface, a CLI, or the API depending on how you want to work with it.
In 2026, OpenAI switched Codex's pricing model from message-based estimates to a direct token-based credit system: usage is metered by actual input, cached input, and output tokens consumed rather than a flat per-message cost, with credits refilling on a rolling 5-hour window. That makes cost more directly tied to task complexity than earlier flat-rate approaches.
Individual tiers run Free, Go at $8/month, Plus at $20/month, and Pro at $100/month (5x capacity) or $200/month (20x capacity). Business teams pay $20 per user/month on annual billing ($25 month-to-month), with custom Enterprise and Education pricing available. A typical working session costs roughly $0.50-$2.00 in credits, with team budgets commonly landing around $100-200 per developer per month.
OpenAI Codex 的主要功能
- Autonomous multi-step coding: Plans, writes, and tests code from a natural-language task description.
- Multiple access points: Available through ChatGPT's Codex interface, a CLI, and the API.
- Token-based credit pricing: Usage is metered by actual token consumption rather than flat per-message pricing.
- Rolling usage window: Credits refill on a rolling 5-hour window rather than a hard monthly reset.
OpenAI Codex 的使用情境
- Delegate a coding task described in plain language to an autonomous agent
- Run coding tasks from the terminal via the Codex CLI
- Integrate autonomous coding capability into a product via the API
- Handle routine, well-specified coding tasks without manual step-by-step guidance