AI Quota Watch

How much AI do you actually get?

Usage limits, plan changes, API prices, intelligence scores and speed for ChatGPT, Claude, Grok and Gemini, in one place. Every number links to where it came from and when we checked it.

Data last reviewed —

Plan quota radar

Bars show each plan's usage relative to that vendor's own 1x baseline. Vendors don't publish absolute message counts, so multiples are not comparable across vendors.

Official read on the vendor's page Reported third-party, not yet confirmed

What changed

Quota and price changes, newest first. Green means users get more, red means less or pricier.

    Reset radar

    Every usage-limit reset we can trace to a post by the vendor, mostly Codex resets announced by Tibo (@thsottiaux). Times are when the post went out; your account may update a little later.

    Resets per month

    Full resets and banked reset cards, last 7 months.

    What time resets land

    Hour of the announcing post, in your time zone. Posts with unknown time are left out.

      Model calculator

      Enter your workload to compare monthly API cost, response time, intelligence and value. Click a column to sort.

      Providers
      Presets:
      Model Intelligence Tokens/s $ in / out per 1M Monthly cost Sec / reply IQ per $ task

      Monthly cost = (input × in-price + output × out-price) × requests × days, list price, no caching or batch discounts. Sec / reply = output tokens ÷ measured tokens/s (excludes time to first token). IQ per $ task = Intelligence Index ÷ Artificial Analysis cost per task.

      Intelligence vs cost per task

      Model radar

      Pick up to four models to compare intelligence, speed, cost efficiency, how far your budget goes, and context size. Each axis is scored 0–100 against the best model in the list. The workload comes from the calculator above.

      Up to 4 models at once.

      Scoring: Intelligence and Speed are divided by the top value. Cost efficiency and Budget reach use a log scale, so 100 is the cheapest model and 0 the most expensive. Budget reach = budget ÷ API cost per request with your calculator workload. A model without a verified API price scores 0 on Budget reach and shows n/a.

      Effort ladder

      The same model at different reasoning efforts: how much smarter each step gets, what it costs, and how many tokens it really burns. Benchmarks come from Artificial Analysis; real usage comes from people who share QuotaPilot data.

      Intelligence vs cost per task, by effort

      Community radar

      Real measurements from people who use these plans every day: speed, how good a model feels, how much a plan really holds, and when limits hit or reset. Rows appear once at least 3 people report the same model or plan in the last 7 days.

      Measured speed

      Median output tokens/s from real sessions, bar shows the middle 50%.

      How it feels

      Average score out of 10. A warning shows when 30% or more of reports say the model got worse.

      Measured plan quota

      Estimated tokens per window: tokens used ÷ share of the limit used. $ is the same tokens at API list price.

      When limits hit and reset

      Reports in the last 30 days by hour of day, in your time zone.

      Share real numbers with QuotaPilot

      QuotaPilot is a free, open-source quota monitor for Codex, Claude Code, Gemini CLI and Grok CLI. It reads your local usage and shares nothing by default. Run the share command to see exactly which anonymous numbers it would send, and send them only if you agree.

      • Measured speed per model from your own sessions
      • Quota used vs. tokens spent, so we can estimate what each plan really holds
      • No prompts, file paths, project names or account ids
      # macOS
brew install --cask dwagwd/tap/quotapilot
# preview, then confirm
quotapilot share
      QuotaPilot on GitHub

      Report a data point

      Takes under a minute. Reports are anonymous: no account, and each browser counts at most a few times a day.

      Get a plan

      Links go to each vendor's official pricing page.

      Sources and method

      We re-check every source by hand and update the date next to it. If an official page cannot be read, we mark the number as Reported.

      Not affiliated with OpenAI, Anthropic, Google or xAI. Limits change without notice; check your account's usage page before you rely on a number. Nothing here is financial advice.