Why there are several: the undated name can move to a newer build when the developer updates it, while a dated snapshot is meant to stay fixed, which can suit pipelines that need repeatable behaviour; an experimental build tests new features. Sellers price each version separately and not every seller offers every version, so check a seller's own docs for the build it serves.
Cost calculator: where to buy, and what it costs you
Pick a use case, a volume and a provider. Monthly figures assume 30 days.
1,500 input and 400 output tokens per request, 30% of input cached.
1101001K10K100K
Spotted a wrong or outdated price? Report it and we fix it at the next update.
A badge with the cheapest price, the seller and the date. It updates every 6 hours and links here.
Markdown[](https://easiapiai.com/m/openai--gpt-5-codex.html)HTML<a href="https://easiapiai.com/m/openai--gpt-5-codex.html"><img src="https://easiapiai.com/badge/openai--gpt-5-codex.svg" alt="GPT-5 Codex API: from $0.81 in / $6.50 out per 1M · Cheaper Inference · 20 Oct 2026" height="40"></a>
Context window
400K400,000 tokens
Cached input
from $0.081per 1M tokens
Last checked
20 Oct, 17:14 UTC
Price history
Cheapest standard price per 1M tokens, recorded every 6 hours since 16 Oct 2026.
OutputInput
No price change has held for 12 hours since 16 Oct 2026.