OpenAI has introduced a new $100 per month ChatGPT Pro plan designed for users who rely heavily on Codex, its AI coding agent built into ChatGPT. The new tier sits between the existing $20 Plus plan and the $200 Pro plan, giving developers a more practical upgrade path without jumping straight to the highest pricing tier.
This move reflects how quickly Codex usage is growing. OpenAI says the tool now has more than 3 million weekly users, with usage increasing rapidly in recent months. The new $100 Pro plan is positioned for developers working on larger or more complex coding tasks who need more consistent access and higher limits than what Plus offers.
The key difference across plans is usage capacity rather than features. The $100 Pro plan includes five times more Codex usage compared to ChatGPT Plus, making it better suited for longer sessions and more demanding workflows. For a limited time through May 31, OpenAI is boosting this further, offering up to 10 times the Codex usage of Plus to help users test more ambitious projects.
Both the $100 and $200 Pro tiers include the same core capabilities. Subscribers get access to OpenAI’s Pro models along with unlimited use of Instant and Thinking models. The real distinction comes down to scale. The $200 plan is built for continuous, high-volume workloads and parallel projects, offering up to 20 times the usage limits of Plus.
OpenAI is also adjusting how usage works on the Plus plan. Instead of enabling long, heavy sessions in a single day, the company is redistributing limits to support more consistent usage throughout the week. This keeps Plus positioned as the better option for steady, everyday coding tasks, while the new $100 tier targets users who need more flexibility and sustained performance.
The updated pricing structure brings OpenAI closer to competitors like Anthropic, which already offers similar mid-tier and high-tier plans. It also highlights a broader shift in AI tooling where pricing is increasingly tied to compute usage and workload intensity rather than just feature access.
