AI plans

$20, $100, or $200 AI plan: how to decide

Start on the $20 plan, move to $100 when you keep hitting limits, and move to $200 when you hit limits again. Use Codex limit resets before you pay for an upgrade.

6 min · 2026-08-01 · Watch on YouTube

The main idea

One day last month I used $300 worth of tokens on my $100 Codex plan. That is three times what I pay for the whole month. It sounds like a reason to upgrade right away, but if you jump straight to a bigger plan you will probably waste money.

My framework is a ladder. Start at the $20 plan. If you do real work, such as agents writing code or research tasks that run for hours, you will hit that limit fast. That is fine. If you hit it consistently and cannot wait for the reset, move to the $100 plan, which gives 5x the use of the $20 plan. If you still hit limits, the $200 plan gives 20x. That is four times the limit for two times the price.

Before you pay for an upgrade, use Codex limit resets. They work even on the $20 plan. On the $100 plan, one reset returns a full 5x weekly budget.

SemiAnalysis bought every ChatGPT and Claude plan, ran coding agents until they hit the weekly limits, and priced the tokens at API rates. The $100 Codex plan gives about $3,500 worth of use if you max it out. The $200 plan gives about $14,000. I think of these plans as rented employees. The goal is to find enough valuable work to keep them busy.

Key points

Chapters

  1. 0:00$300 of tokens in one day on a $100 plan
  2. 0:51Rung 1: why most people start at $20
  3. 1:35The full ladder: 4x the limits for 2x the price
  4. 3:05What each plan is worth in tokens
Free live lesson · Nov 24Stop Measuring AI by Tokens. Measure What It EarnsOne hour, live, with Saurav. Bring your own bill. Free.
Save my free seat →
Get new videos and tips by emailOne short email when there is something new. Unsubscribe any time.

More CloudYeti videos