OpenAI's $100 Pro signal

OpenAI just inserted a new price point between $20 and $200, and the reason is not casual chat. On April 9, OpenAI said its new $100-a-month ChatGPT Pro tier is built to support heavier use of Codex, its coding agent inside ChatGPT. (community.openai.com, theverge.com) The new plan gives 5 times more Codex usage than ChatGPT Plus, which still costs $20 a month. OpenAI also said that through May 31, 2026, $100 Pro subscribers will get up to 10 times the Codex usage of Plus as a launch promotion. (community.openai.com) Codex is not a better autocomplete bar. OpenAI describes it as a cloud software engineering agent that can work on many tasks in parallel, which means people use it for longer sessions that look more like assigning work to a junior engineer than asking a chatbot one question. (openai.com) That changes what a subscription price is buying. OpenAI’s own help page now frames Plus as “for lighter use” and the $100 Pro plan as “built for real projects,” with higher allowances for Codex and Deep Research. (help.openai.com) The old ladder had a strange gap. Public reporting says OpenAI had free, $8 Go, $20 Plus, and then a $200 Pro option, so developers who had outgrown Plus were jumping straight to a plan priced for the heaviest users. (techcrunch.com, tech.yahoo.com) A $100 middle tier usually appears when a company has found a group of users with a repeatable job to do. In this case, the repeatable job is writing, editing, testing, and shipping code often enough that more Codex capacity is worth $80 more than Plus but not $180 more. (theverge.com, venturebeat.com) OpenAI’s own product pages point the same way. The company says GPT-5.4 improves “developer workflows” across ChatGPT, the application programming interface, and Codex, and its Codex pricing page says the tool is now included across ChatGPT plans from Free through Enterprise. (openai.com, developers.openai.com) Once a lab starts charging specifically for coding-agent usage, the bottlenecks stop being only model intelligence. The valuable work shifts toward evaluation systems that can tell whether generated code actually works, latency cuts that make long coding sessions feel usable, and product design that helps developers steer an agent without fighting it. (openai.com, openai.com) That is why this pricing move doubles as a hiring signal. A researcher who can raise coding benchmark scores, a systems engineer who can make Codex faster, or a product builder who can make agent handoffs cleaner is now tied more directly to subscription revenue than in a chat-only business. (openai.com, help.openai.com) The short version is that OpenAI is no longer pricing ChatGPT mainly as a smarter search box. It is pricing part of ChatGPT as a software workbench, and companies that make money that way usually start valuing people who improve the whole workflow, not just the model at the center of it. (community.openai.com, openai.com)