OpenAI is making it a lot easier for teams to try Codex without committing to high upfront costs. Starting now, companies on ChatGPT Business and Enterprise can add Codex-only seats on a pay-as-you-go basis, so small groups can run pilots, automate a few high‑value workflows, and then scale if it actually proves useful. Instead of paying a fixed seat fee, these Codex-only seats are billed purely on token usage, with no rate limits, which means finance and engineering leaders get a much clearer line of sight from “who’s using Codex for what” to “what did that actually cost us this month.”

This change also comes with a price tweak on the broader ChatGPT side. Standard ChatGPT Business seats, which still include Codex usage but with limits, are dropping from $25 dollars to $20 per user per month on annual billing, making full ChatGPT access a bit more accessible for teams that want general AI plus some coding help. The idea is pretty straightforward: if you just need Codex deeply embedded in engineering workflows, you can buy that à la carte; if you want ChatGPT for everyone, that path is cheaper too.

OpenAI is clearly betting on Codex as a serious team tool, not just a solo developer assistant. According to the company, more than 9 million paying business users rely on ChatGPT at work, and over 2 million builders are using Codex every week, with Codex usage inside ChatGPT Business and Enterprise growing 6x since January. Teams at companies like Notion, Ramp, Braintrust, and Wasmer are already using it to accelerate code generation, standardize workflows, and move from scattered “AI experiments” to more systematic engineering automation.

To get started, OpenAI is pushing the Codex desktop app for macOS and Windows, which plugs straight into existing tools and now ties into newer features like Plugins and Automations, so teams can hook Codex into their internal systems with less glue work. There is also a short‑term adoption perk: eligible ChatGPT Business workspaces can get $100 in credits for each new Codex-only team member who actually starts using the product, capped at $500 per team. For a small engineering squad, that’s enough to meaningfully test Codex in real workflows—code review helpers, internal tool scripts, or repetitive ticket handling—before deciding whether to roll it out more widely.