Jul, 134 min to read
Frontend
News · OpenAI splits Codex billing from ChatGPT seats with token-based pricing
Jul, 94 min to read
FrontendOpenAI splits Codex billing from ChatGPT seats with token-based pricing
A pay-as-you-go Codex seat, token metering, and a quiet reversal for Business plans reshape how teams pay to put a coding agent in front of engineers.
What OpenAI actually changed
The core move is a billing split. Teams on ChatGPT Business and Enterprise can now add Codex-only seats that carry no fixed seat fee and are billed on token consumption instead. Those Codex-only seats also carry no rate limits, which is the trade OpenAI is making: you stop paying for a seat you may barely use, and start paying for exactly the tokens you burn.
That distinguishes two paths. Standard ChatGPT Business seats still include Codex usage but keep usage limits, and OpenAI lowered that annual price from $25 to $20 per seat. Codex-only seats remove the ceiling and the flat fee, trading predictability for metered cost visibility. OpenAI frames this as clarity — 'a clearer view of how usage turns into spend' — which matters when a coding agent's cost scales with how hard engineers lean on it.
The reversal buried in the update note
The most revealing detail is the June 24 update stapled to the top of an April announcement: new Codex pay-as-you-go seats are no longer available for Business plans, though existing seats keep working. So the flexible-pricing option that headlines this post was withdrawn from one of its two named audiences within roughly three months.
Read plainly, the metered model survives for Enterprise and grandfathered Business customers but is closed to new Business signups. That suggests token-based, unlimited Codex access behaved differently than a flat seat fee at Business scale — the kind of pattern where uncapped consumption on a coding agent produces bills that don't fit the plan it was sold under. OpenAI didn't explain the reversal, but the timing tells engineering leads to plan around Enterprise, not Business, if metered Codex is the goal.
The adoption push aimed at frontend and product engineering
OpenAI is pairing pricing changes with distribution. The recommended entry point is the Codex app for macOS and Windows, and two new capabilities — Plugins and Automations — are positioned to connect Codex to systems teams already run. For frontend teams, that connection layer is where a coding agent stops being a chat window and starts touching the build tools, design systems, and CI paths where day-to-day work lives.
There is also a direct financial nudge: eligible Business workspaces can receive $100 in credits per new Codex-only member who joins and starts using Codex, capped at $500 per team, for a limited time. The stated growth backdrop is a 6x increase in Codex users inside ChatGPT Business and Enterprise since January, against more than 2 million weekly Codex builders and 9 million paying business users overall.
What a metered coding agent means for how frontend teams pilot AI
The design goal OpenAI states is letting small groups 'begin pilots, prove value in a few critical workflows, and easily expand from there.' Token billing suits that shape: a two-person pilot on a component-library migration or a test-generation workflow can prove value without committing a whole team to seats. But the June reversal complicates the promise — the frictionless pilot path is now the Enterprise path, and Business teams that didn't already claim seats can't start there.
The practical implication is to treat Codex cost as a usage line, not a headcount line, and to instrument it accordingly. Named customers like Notion, Ramp, Braintrust, and Wasmer are cited for faster execution and more repeatable workflows; the repeatability is what makes token spend forecastable. For a frontend team, the question isn't whether unlimited Codex access helps — it's whether your workflows are structured enough that metered usage stays legible before an uncapped agent quietly becomes your largest AI invoice.
Find this story relevant to you?
Contact us to find a unique solution
Need an AI engineering partner that can actually build?
We help businesses integrate AI, build AI-powered products, automate high-value workflows, and modernize the software systems behind them.
Related reading
More analysis around product delivery, operational AI, and the systems work that makes deployment hold up in reality.