OpenAI reshuffles ChatGPT lineup

OpenAI is narrowing what most ChatGPT users see and charging more for the people who want its heaviest coding tools. (help.openai.com) The company said models including GPT-4o, GPT-4.1, GPT-4.1 mini, OpenAI o4-mini, and GPT-5 Instant and Thinking were retired from ChatGPT on February 13, 2026, while application programming interface access stayed unchanged. GPT-5.3 Instant is now the default for logged-in users, and GPT-5.4 Thinking is positioned for harder research, spreadsheet, and coding work. (help.openai.com) (openai.com) OpenAI also added a new $100-a-month ChatGPT Pro plan on April 9 that sits between its $20 Plus plan and $200 Pro plan. Engadget reported the new tier includes the same core tools and models as the $200 plan but with lower usage limits, and five times more Codex usage than Plus. (engadget.com) That matters because ChatGPT is shifting from a single chatbot into a tiered software bundle for work. OpenAI’s own product posts now describe Codex as a “command center” for multiple coding agents and GPT-5.4 as a model built for professional tasks across documents, spreadsheets, browsers, and software tools. (openai.com 1) (openai.com 2) The lineup change also simplifies a product catalog that had grown crowded with overlapping names and modes. Instead of asking users to pick among many legacy models, OpenAI is steering mainstream customers toward one default model and reserving higher-capability workflows for paid tiers. (help.openai.com) (cnbc.com) OpenAI has been building the coding side of that strategy for months. It introduced the Codex desktop app on February 2, said on March 4 that it had expanded to Windows, and described the app as a place to run multiple agents in parallel on long-running software tasks. (openai.com) On March 5, OpenAI said GPT-5.4 would ship in ChatGPT, Codex, and the application programming interface, and called it its first general-purpose model with native computer-use abilities. The company said the model can work across tools and hold up to 1 million tokens of context, which is a way to keep a very large amount of text and instructions in memory at once. (openai.com) Reporting from CNBC on March 19 said OpenAI plans to combine its browser, ChatGPT app, and Codex app into one desktop “super app” led by applications chief Fidji Simo with help from President Greg Brockman. CNBC said the goal is to reduce fragmentation and push harder into high-productivity use cases. (cnbc.com) TestingCatalog reported on April 11 that code references inside Codex point to a broader app that could absorb ChatGPT and the Atlas browser, plus a “Scratchpad” feature for launching multiple Codex tasks in parallel. OpenAI has not publicly confirmed those specific features, but they fit the direction the company and CNBC have already outlined. (testingcatalog.com) (cnbc.com) The result is a cleaner message and a sharper paywall: everyday chat goes to the default model, while advanced coding, longer-running agents, and heavier usage move deeper into premium plans. That makes OpenAI look less like a menu of models and more like a workplace software company selling access by intensity. (help.openai.com) (engadget.com) (openai.com)