AI tools: OpenAI and Meta

OpenAI and Meta both used this week to make a messy part of the artificial intelligence business easier to see: not the chat box, but the meter running behind it. OpenAI rewrote how coding credits work for business customers on April 2, 2026, and Meta unveiled a new in-house model line called Muse on April 8, 2026. (help.openai.com) (about.fb.com) OpenAI’s change was about Codex, its coding-focused assistant, which had been billed partly like a bundle of messages. The new rate card says new and existing ChatGPT Business customers, plus new ChatGPT Enterprise customers, now pay based on token usage, which is closer to paying for electricity by the kilowatt-hour than by the number of times you flip the switch. (help.openai.com) (developers.openai.com) That sounds small until you look at who gets moved and who does not. OpenAI says existing Plus, Pro, and Enterprise or Education customers stay on the legacy message-based Codex pricing for now, while Business workspaces are already on the new token model. (help.openai.com) OpenAI also changed the shape of its business plan itself on April 2, 2026. The company added a Codex-only seat for people who need coding access without a full ChatGPT seat, and it cut the price of standard ChatGPT Business seats by 5 United States dollars per month before regional adjustments. (help.openai.com 1) (help.openai.com 2) The release notes show where OpenAI thinks those seats will be used. On March 27, 2026, ChatGPT Business got updated Box, Notion, Linear, and Dropbox apps with new actions, including write capabilities where the app supports them, which turns the assistant from a search window into something closer to a junior employee that can edit files and push work into other software. (help.openai.com) OpenAI even attached a short-term sales push to the new setup. A help article says eligible ChatGPT Business workspaces can earn 100 dollars in promotional credits for each newly added Codex seat that sends its first Codex message, up to 500 dollars total, starting April 2, 2026. (help.openai.com) Meta’s announcement came from the other direction. Instead of clarifying billing, Meta introduced a new family of large language models called Muse and said the first model, Muse Spark, already powers the Meta AI app and website. (about.fb.com) Meta says Muse Spark will roll out to WhatsApp, Instagram, Facebook, Messenger, and its artificial intelligence glasses in the coming weeks. It also says the model will later unlock features that cite recommendations and content people share across Instagram, Facebook, and Threads, which ties the model directly to Meta’s social graph rather than treating it like a generic chatbot floating above the apps. (about.fb.com) That is the link between these two announcements. OpenAI is making it easier for companies to buy artificial intelligence work in smaller, more measurable pieces inside office software, while Meta is building artificial intelligence directly into consumer apps where recommendations, posts, messages, and product discovery already live. (help.openai.com) (about.fb.com) For anyone in marketing, product, or operations, the practical shift is that “using artificial intelligence” is no longer one subscription decision. OpenAI is separating chat access from coding usage with seats and token meters, and Meta is turning its own apps into the distribution layer for new model features, with a private preview application programming interface for selected partners on top. (help.openai.com) (about.fb.com)