OpenAI updates GPTs

OpenAI has expanded the models available inside custom ChatGPT bots that call outside services, adding GPT-4o and GPT-4.1 on the web for paid users. (help.openai.com) Those custom bots, called GPTs, can be wired to external application programming interfaces through “actions,” which let them fetch data or trigger tasks in other software. OpenAI’s help documentation says builders need an OpenAPI schema plus authentication such as an application programming interface key or OAuth to connect those services. (help.openai.com) The rollout listed Plus, Pro, and Team users on the web first, while Enterprise and education availability was marked “coming soon” in the release notes. OpenAI renamed ChatGPT Team to ChatGPT Business on August 29, 2025, according to a separate billing article. (help.openai.com 1) (help.openai.com 2) OpenAI also published a Codex rate card that breaks usage into credits per 1 million input, cached input, and output tokens instead of per-message estimates. The company said that token-based pricing took effect April 2, 2026 for new and existing Plus, Pro, and Business customers, plus new Enterprise customers. (help.openai.com) The new table shows different credit costs by model, with GPT-5.4 priced at 62.50 credits per 1 million input tokens and 375 credits per 1 million output tokens, while GPT-5.1-Codex-mini is listed at 6.25 input credits and 50 output credits. OpenAI also says Fast mode uses twice as many credits and that code review uses GPT-5.3-Codex. (help.openai.com) That pricing change sits alongside a broader shift in ChatGPT plans toward metered usage for advanced features. OpenAI’s flexible-pricing documentation says credits can be used for Deep Research, Thinking models, image generation, Advanced Voice, and Codex, and that Business and Enterprise plans now include a Codex-only seat type introduced April 2, 2026. (help.openai.com 1) (help.openai.com 2) The model update also lands after OpenAI spent the past year separating what is available in the application programming interface from what appears inside ChatGPT. When OpenAI introduced GPT-4.1 on April 14, 2025, it said the model family would be available through the application programming interface, while ChatGPT would absorb some of its improvements separately. (openai.com) In practice, the latest help pages point to a more segmented OpenAI product lineup: consumer subscriptions get more model choice inside GPTs, while businesses get finer controls over seats, credits, and workspace spending. The next step is the delayed Enterprise and education rollout that OpenAI has already signaled in its release notes. (help.openai.com 1) (help.openai.com 2)