OpenAI rarely names a competitor by name in official announcements. But in April 2026, the company did just that: a letter to journalists stated that Codex "delivers more coding capabilities per dollar compared to Claude Code." The new $100 per month plan is publicly positioned as a direct response to Anthropic.
What $100 gets you, and what it doesn't
The new plan closes an obvious gap in OpenAI's lineup: previously there was nothing between $20 (Plus) and $200 (Pro). Now there are five tiers for individuals: Free, Go, Plus, and two Pro levels.
For $100 per month, users get access to the same models and features as the more expensive $200 subscription — but with lower limits. Specifically for Codex, this means five times more than Plus offers, while the $200 plan provides 20 times more. The plan is positioned as a solution for "long, resource-intensive coding sessions."
The number of Codex messages you can send within limits depends on the size and complexity of your tasks.
— OpenAI, FAQ for the new plan
One detail worth noting that's easy to miss: current limits for Pro 5x are temporarily doubled — until May 31, 2026. After that date, OpenAI may reduce them — and formally won't break any promises.
Why now
According to The New Stack, OpenAI reports that over 3 million people weekly use Codex — a 5x increase in three months, with growth exceeding 70% monthly. There's demand. But Anthropic managed to offer its $100 plan first (Max 5x for Claude Code), and some developers who lacked Plus limits switched there.
The new plan is an attempt to win them back or stop the exodus. According to CNBC, Anthropic's subscription structure is nearly a mirror image: it has Free and four paid tiers with similar $100/$200 steps at the top.
Practical breakdown
- Plus ($20) — suitable for occasional Codex use and general ChatGPT
- Pro 5x ($100) — for developers who run agent tasks daily but don't need parallel projects 24/7
- Pro 20x ($200) — for teams or intensive continuous workloads
There's also an exclusive: the GPT-5.3-Codex-Spark model in research preview is available only for both Pro tiers — it runs on specialized low-latency hardware, and its limit may vary depending on load.
What's next
If after May 31 OpenAI does reduce the temporarily boosted limits — the developer community's reaction will be a real test of whether the $100 plan is truly competitive with Claude Code, or merely appears so at launch.
