OpenAI has introduced a new $100-per-month Pro tier for ChatGPT, positioning it directly between the $20 Plus plan and the existing $200 Pro subscription. Announced on April 9, 2026, the tier is clearly aimed at users who find the Plus limits too restrictive for extended coding sessions but do not need the full capacity of the higher-priced option. It offers five times the Codex usage of the Plus plan, along with access to the same advanced model suite available on the $200 tier.
The pricing structure for ChatGPT now spans six levels, ranging from a free ad-supported account and an $8 Go plan up to custom enterprise contracts. The new $100 Pro sits in the middle, targeting what the company calls “longer, high-effort Codex sessions.” For context, the $200 Pro provides 20 times the Codex usage of Plus, making the $100 tier roughly one-quarter as generous in volume while matching it in model access. Both Pro plans include the exclusive GPT-5.4 Pro model, unlimited GPT-5.4 Instant and GPT-5.4 Thinking, plus all other features from the top tier.
A launch promotion sweetens the deal temporarily: until May 31, 2026, new $100 subscribers receive ten times the Codex usage of Plus. After that date, the standard five-times limit applies. At the same time, OpenAI adjusted the Plus plan’s Codex allocation to better suit lighter, day-to-day tasks, nudging heavier users toward the new mid-tier.
The timing reflects rapid growth in Codex adoption. On April 8, OpenAI noted that weekly Codex users had reached three million, up from two million just weeks earlier and representing a fivefold increase over three months with 70 percent month-over-month growth. This surge coincides with the February 2026 launch of a dedicated macOS Codex app designed for more agentic workflows — multi-step, background-running tasks that can stretch across hours rather than delivering quick line-by-line suggestions. Such extended sessions quickly exhaust Plus limits, creating the exact gap the $100 plan is meant to fill.
OpenAI made little effort to hide the competitive intent. The price matches Anthropic’s $100 Claude Max tier, which offers elevated limits for its own agentic coding tool, Claude Code. That product has become one of Anthropic’s fastest-growing segments, reportedly generating around $2.5 billion in annualized revenue by early 2026. The week before OpenAI’s announcement, Anthropic tightened rules by banning third-party agents from routing usage through its Pro and Max plans, forcing extra charges for heavy external framework use. OpenAI took the opposite approach, expanding availability at the same price point and doubling it temporarily for launch.
Behind the scenes, the move fits into OpenAI’s broader commercial momentum. The company recently closed a $122 billion funding round at an $852 billion valuation, backed by SoftBank, NVIDIA, Amazon, and even retail investors — steps many interpret as preparation for a possible IPO later in 2026. OpenAI is currently generating roughly $2 billion in monthly revenue and counts more than 50 million paid subscribers. Filling the pricing gap between $20 and $200 was a logical step to capture developers who want serious coding capacity without jumping straight to enterprise-level spending.
At its core, the $100 plan is a bet on the future of AI-assisted development. GPT-5.4, released in March 2026, brought native computer-use capabilities directly into Codex, enabling autonomous agents that can navigate interfaces, manage files, and orchestrate complex workflows. The new tier prices exactly that shift — moving beyond simple prompting toward software that can act independently over extended periods.
Whether the plan will pull meaningful subscribers away from Claude Max remains to be seen. Usage metrics in the coming quarters will show if the combination of higher limits, temporary doubling, and competitive pricing is enough to sway heavy users in a market that is growing quickly but still fragmented between the two leading players. For now, it gives power users a more graduated upgrade path while OpenAI continues to push the boundaries of what agentic coding tools can actually deliver in everyday development work.
