The cheapest way to use OpenAI Codex in 2026

Codex is excellent, but paying full price for every run adds up. Here's the cheapest honest way to run it.

Free to start

AnyModel uses pay-per-token billing with no subscription or minimums. Link Telegram to receive 5,000,000 free tokens; registration itself does not include a token grant.

One-command setup

bash <(curl -fsSL "https://anymodel.org/i?tool=codex") <YOUR_API_KEY>

Pick the right tier and pay less

You don't have to run the most expensive model for everything:

• GPT-5.2 Codex for hard, autonomous coding.
• GPT-5.1 Codex — cheaper, handles most daily work.
• Or switch the model id to a budget option like DeepSeek or GLM for bulk tasks.

Same endpoint, same key — just change the model. See the full price list or compare two side by side on any comparison page.

Pay per token, not per month

No subscription, no minimums. You're billed only for the tokens you actually use. Start free.