ChatGPT Pro $100 promises 5x Plus for Codex. We measured 2.6x
Updated October 9, 2026 · first published October 9, 2026
OpenAI sells ChatGPT Pro $100 as 5x the Codex usage of Plus. We drained 6 full Pro $100 weekly windows between September 27 and October 8, 2026, and priced every one of them from Codex's own logs. The median window held $175 of API-priced Codex. A Plus week holds about $67. That is 2.6x, roughly half of what the label says.
- Label: 5x Plus, about $335 of API value a week.
- Measured: $100 to $234 per full window, median $175.
- Best window: 3.5x. Worst: 1.5x. Not one window reached 4x.
The math
The unit we use is API-equivalent dollars: every request's input, cached input and output tokens, priced at OpenAI's API list price for the model that served it. Plus is the baseline. Across 27 Plus weekly windows since August, a full Plus week priced at a median $67 (the full series is on our Codex usage limits page).
| API value per week | Multiple of Plus | |
|---|---|---|
| Plus ($20) | $67 | 1.0x |
| Pro $100, as advertised | $335 | 5.0x |
| Pro $100, measured median | $175 | 2.6x |
| Pro $200 after Oct 30, at the label (10x) | $670 | 10x |
| Pro $200 after Oct 30, at our measured ratio | about $351 | about 5.2x |
Per dollar of monthly price, a Plus week holds $3.35 of Codex. A Pro $100 week at our measured median holds $1.75. You pay 5x for 2.6x: the upgrade buys throughput, not a better unit price.
Why the gap?
Three things could explain it, and we cannot fully separate them from the outside.
- The gauge does not track API prices. OpenAI meters Codex in its own units. If a cheap model like Luna draws on the gauge at a higher rate than its API price, dollars understate the quota used. Luna was 49% of our tokens but only 8% of the dollars.
- Heavier sessions. Our Pro windows ran longer, more cache-heavy agent loops than our Plus weeks. Window 4 (Sep 30) was the cheapest at $100, almost entirely long GPT-6 Sol sessions.
- The label is a ceiling, not a promise. OpenAI describes limits as approximate and dependent on task size. Others see the same ratio: one forum user who moved up from Pro $100 reported roughly half the effective usage they expected.
Measure your own multiplier in five minutes
Codex writes a JSONL log of every session under ~/.codex/sessions/. Each request records cumulative input, cached input and output tokens, plus the weekly gauge (used_percent) and its resets_at time. We priced every request's token delta at the API list price of the model it ran on and grouped requests by weekly window. Dollar figures are API-equivalent value, not money we paid.
- Pick a weekly window you used to 100%.
- Sum the API value of every request inside it. That is your dollars per full window.
- Do the same for a Plus week, or use our $67.
- Divide. If your gauge stopped at, say, 80%, divide the dollars by 0.8 first.
The raw numbers for all eight of our Pro $100 windows are in codex-pro-100-windows.csv: start, end, hours, percent used, API value, value per full window and multiple of Plus.
Is Pro $100 still worth it?
For heavy Codex users, yes, but because of resets more than the multiplier. In 12 days we got eight windows instead of two, and the resets were worth about 4x the plan itself. On the multiplier alone, Pro $100 is 2.6x Plus for 5x the price. And from October 30, Pro $200 drops to 10x, which at our measured ratio is about $351 a week.
Sources
- OpenAI: Codex plans and usage limits
- VentureBeat: ChatGPT Pro $100 tier with 5x Codex limits
- OpenAI Community: Pro $200 giving roughly half the effective usage of Pro $100
- OpenAI API pricing
Related
- Codex resets turned $40 into $1,211
- 8 Codex weekly windows in 12 days
- OpenAI halved ChatGPT Pro $200
- How much Codex ChatGPT Pro includes
- When to redeem a Codex banked reset
- Claude Max 20x vs ChatGPT Pro
Want this applied to your stack? Bring the provider bills, gateway logs, and top workflows; we will map the cost drivers and savings path. Book a free audit →