OpenAI launches free Codex tier, undercutting Claude Code’s $20/month plan

Coding agents just got a lot cheaper to try. That is the real news here. OpenAI’s Codex pricing page now shows Codex included in ChatGPT Free, alongside paid tiers like Go, Plus, and Pro, which means the floor price to start using its coding agent has dropped to $0. (developers.openai.com) Anthropic still lists Claude Code as part of Claude Pro, which starts at $20 a month if billed monthly. (claude.com) ### What changed today? The clearest signal is the pricing page itself. OpenAI now says “Codex is included in your ChatGPT Free, Go, Plus, Pro, Business, Edu, or Enterprise plan,” and labels Free as “$0/month” for quick coding tasks. (developers.openai.com) That is a straightforward shift from Codex being mostly framed around paid ChatGPT access when the Codex app launched earlier this year. ### Wasn’t Codex already available before? Yes — but not like this. When OpenAI introduced the Codex app on February 2, 2026, it said Free and Go access was only available “for a limited time,” while standard inclusion centered on Plus, Pro, Business, Enterprise, and Edu users. (developers.openai.com) The current pricing page and Help Center now present Free as a listed access tier, not just a temporary promo footnote. That is the important difference. ### What exactly is Codex here? This is not the old autocomplete-era Codex brand from years ago. OpenAI is using Codex as a full coding agent — one that can understand a repo, work across files, run commands in a sandbox, and help turn tasks into reviewed pull requests. (developers.openai.com) The company pitches it as an “AI coding partner” that works across the web app, CLI, IDE extension, and other surfaces tied to a ChatGPT account. ### How does Claude Code compare? Claude Code is also an agentic coding tool, not just a chatbot that happens to write snippets. Anthropic describes it as a system that reads a codebase, makes changes across files, runs tests, and delivers committed code. (openai.com) On Anthropic’s current pricing page, Claude Code is included with Pro, and Pro costs $20 monthly or $17 a month if paid annually. ### Why does the $0 versus $20 gap matter? Because entry price shapes experimentation. A lot of developers are not running giant autonomous coding workflows all day. They are poking at a side project, testing an internal tool, or seeing whether an agent can handle bug fixes without breaking things. (openai.com) For that kind of lightweight use, a $20 minimum acts like friction. A free tier removes that friction. ### Does free mean unlimited? No — and that is the catch. OpenAI’s pricing page frames the Free tier as a way to “explore Codex capabilities on quick coding tasks,” and its Help Center says usage depends on model, task size, and whether work runs locally or in the cloud. (anthropic.com) So this is better understood as a low-cost on-ramp, not a blank check for heavy agent use. ### Is Anthropic standing still? Not really. Anthropic said this week that it is doubling Claude Code’s five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans, and removing peak-hours limit reductions for Pro and Max. (developers.openai.com) So Anthropic’s answer looks less like “go cheaper” and more like “make paid usage stronger.” ### Who benefits first? Hobbyists do. Students do. Small teams doing evals do. Also companies that want to test agent workflows before they decide whether these tools belong in CI, code review, or internal engineering ops. Basically, OpenAI just made “let’s try a coding agent” a much easier sentence to say yes to. (developers.openai.com) ### Bottom line This is a pricing move disguised as a product update. OpenAI did not just make Codex cheaper — it erased the starting price. That does not guarantee Codex wins on quality or serious developer loyalty. But it absolutely changes the first decision point, and in software tools, that first decision point matters a lot. (anthropic.com) (developers.openai.com)