OpenAI temporarily grants GPT‑5.5 access to 8,000 developers and adds purchasable ChatGPT credits

ChatGPT pricing just got more complicated — and more enterprise-shaped. OpenAI has started selling flexible credits inside ChatGPT plans, so users can buy extra usage when they hit built-in limits. At almost the same moment, it handed more than 8,000 developers a temporary 10x Codex usage boost after a GPT‑5.5 event filled up. Those two moves look separate, but basically they point the same way: ChatGPT is becoming a metered work system, not just a monthly app. (help.openai.com) ### What actually changed on May 5? OpenAI’s new help docs say flexible pricing went live on May 5, 2026 for ChatGPT plans, letting users purchase credits once they run past included usage. Those credits are portable across supported features instead of being locked to one tool. Right now that means Codex for Plus and Pro subscribers, (help.openai.com) 26, with purchased Sora credits rolling into the broader ChatGPT credit system. (help.openai.com) ### Why does that matter? Because subscriptions used to hide the meter. You paid for Plus or Pro and mostly thought in terms of access, not consumption. Credits change the mental model. Now heavy usage has a visible incremental cost, which is a much better fit for coding agents and media generation — products where one user can burn far more compute than another in a single afternoon. (help.openai.com) ### What was the GPT‑5.5 giveaway? OpenAI also emailed people who applied to an invite-only GPT‑5.5 event and couldn’t all be accommodated. More than 8,000 developers had expressed interest within 24 hours, and OpenAI gave those applicants a consolation perk: 10x higher Codex rate limits on their personal ChatGPT accounts through June(help.openai.com) Codex for a month and see what sticks. (venturebeat.com) ### Why tie GPT‑5.5 buzz to Codex? Because coding is where extra usage turns into habit fast. If a developer gets used to asking Codex to inspect a repo, draft changes, or run repeated iterations, higher limits are not just a perk — they are a workflow unlock. OpenAI’s own Codex help page now frames Codex as included across paid ChatGPT tiers, which makes the giveaway feel less like a one-off freebie and more like a funnel into ongoing paid usage. (help.openai.com) ### What changed for admins? OpenAI’s enterprise and education docs now get much more explicit about model availability and limits. There are pages for Business and for Enterprise/Edu that list which models are available, which older ones were retired from ChatGPT, and how usage caps work. Separate admin docs also describe role-based usage limits, default workspace limits, a(help.openai.com)ools across teams. (help.openai.com) ### Why is governance suddenly the point? Because once usage is metered, somebody owns the budget. That changes the conversation from “should employees have access?” to “who gets how much access, to which models, with what cap?” Coding agents especially push companies into this territory, since the most valuable users are often the ones most likely to blow through limits. OpenAI’s docs are starting to look like the control plane for that problem. (help.openai.com) ### Is this really about Sora too? Yes, but maybe less than it first appears. The credits are shared across supported features, and OpenAI explicitly says credits bought in one place can be used in another. But the bigger signal is packaging, not video. OpenAI is building one spend bucket that can follow the user across tools, which is exactly how a platform starts to behave like internal infrastructure. (help.openai.com) ### Bottom line? The headline is not just that 8,000 developers got a temporary Codex bump. It is that OpenAI is teaching users and admins to think in quotas, overages, and shared pools. That is a small pricing tweak on paper — but in practice it is how consumer chat software turns into managed enterprise compute. (help.openai.com)