OpenAI reshapes access and pricing
OpenAI removed newer GPT‑5.3/5.4 models from ChatGPT while keeping GPT‑4o available to Business, Enterprise and Edu customers in some settings, signalling tighter product tiers and access controls. The company also published a Codex rate card and appears to be packaging a $100 Pro plan tied to Codex usage, a move aimed at separating consumer and paid developer offerings (help.openai.com) (help.openai.com) (economictimes.indiatimes.com).
OpenAI just pulled a strange move: it retired GPT-5.3 and GPT-5.4 inside ChatGPT, even while those names still show up in OpenAI’s model notes and coding products. The help page says the change took effect on February 13, 2026, and that application programming interface access was unchanged. (help.openai.com) That split matters because ChatGPT and the application programming interface are now being treated like two different stores selling from the same warehouse. The consumer chat app lost access to several newer models, while developers can still reach retired ChatGPT models through the application programming interface. (help.openai.com 1) (help.openai.com 2) The cleanup was broader than two model names. OpenAI’s release notes say GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and GPT-5 Thinking were all retired from ChatGPT on February 13, 2026. (help.openai.com 1) (help.openai.com 2) Business customers got a short grace period instead of a full exemption. OpenAI said ChatGPT Business, Enterprise, and Education customers could keep using GPT-4o inside custom versions of ChatGPT until April 3, 2026, after which GPT-4o would be fully retired across all plans. (help.openai.com 1) (help.openai.com 2) At the same time, OpenAI changed how Codex is billed. On April 2, 2026, the company said Codex pricing would move from per-message charges to application programming interface token pricing for new and existing Plus, Pro, and Business plans, plus new Enterprise plans. (help.openai.com) A token is the meter that counts how much text a model reads and writes, like paying for electricity by kilowatt-hour instead of paying per light switch flip. Moving Codex to token pricing makes coding usage easier to separate from ordinary ChatGPT conversations. (help.openai.com) (developers.openai.com) OpenAI’s developer pricing page makes the separation even clearer. It says some Codex paths offer delayed access to new models like GPT-5.3-Codex and GPT-5.3-Codex-Spark, which means the newest coding models are being used as a perk to sort customers by plan and workflow. (developers.openai.com) Then came the new price anchor. The Economic Times reported on April 10, 2026 that OpenAI introduced a $100 Pro plan for Codex with five times the Codex usage of the $20 Plus tier, while Anthropic already sells Claude Max plans at $100 and $200 per month. (economictimes.indiatimes.com) That report also said OpenAI is tightening what a normal subscription covers. Third-party integrations such as OpenClaw would no longer count under standard subscription limits, and usage through those tools would move to a separate pay-as-you-go model. (economictimes.indiatimes.com) Put together, the pattern is simple: fewer flagship models inside plain ChatGPT, more metered pricing around coding, and clearer fences between casual users, workplace buyers, and developers. OpenAI is no longer acting like one subscription should cover chat, software building, and heavy agent use all at once. (help.openai.com) (help.openai.com) (developers.openai.com)
