Free tool
What is one percent of your plan worth? The tokens behind 1% of the 5-hour session and of the week, plan by plan, measured by UsageMeter and set next to what each plan promises.
A plan needs at least 5 accounts before this page publishes a median. Until then it shows what each plan promises and how the measurement works, and nothing more.
Accounts measured in the last 30 days. Read on November 3, 2026.
Standard seats get more than Pro, and Premium seats get 5x Standard. Anthropic gives no multiplier against Pro, and the two kinds of seat cannot be told apart here.
The barometer grows with measurements. UsageMeter and UsageMeterLink add yours, counted with the others with no name shown, and show your own limits live.
Only what Anthropic’s pricing page says, checked on October 30, 2026. A promise is a ratio to Pro, not a number of tokens: Anthropic does not publish the size of any limit.
One percent is one point of the gauge Claude shows in Settings > Usage. The figure is how many tokens of Claude Code that point was worth, as the median of the accounts on the plan. “× Pro” divides it by Pro’s figure: set it next to the multiplier the plan promises.
A ratio close to the promise is consistent with it. A ratio far from it is a question, not a finding: the gap can come from the sample, the mix of models, a promotion or the way the gauge counts.
UsageMeterLink reads the tokens of every Claude Code turn on a computer. Between two readings of Claude’s own gauge, UsageMeter divides the points the gauge moved by the tokens of those turns, one model family at a time, and keeps the low edge of the results, since usage it cannot see only makes the gauge move faster. That gives a ceiling for 100% of the limit; one percent is a hundredth of it.
The page takes the latest value of every account measured in the last 30 days, groups the accounts by the plan Claude reported for them, and publishes the median of a plan only when at least 5 accounts stand behind it. Tokens are input, output, and cache reads and writes.
Each account counts once, whatever its volume. No account, address or exact count is shown: only the range of accounts.
It measures Claude Code turns. Chats on claude.ai bring no tokens to UsageMeter, and usage it cannot see moves the gauge without tokens, which tends to pull a figure down. The method keeps the low edge of its results to limit that, not to remove it.
The data are young and thin: that is why most plans show no figure yet, and why a figure that does appear is a median of a few accounts.
A promotion applied to some accounts only, a plan change, a different mix of models or a change in how Claude counts will move a median. This page measures; it never says why.
Nothing here says that Anthropic reduced a limit. When a limit changes, Anthropic announces it, and our dated timeline records it.
One email each time Anthropic changes the usage limits, taken from our dated timeline. Nothing else, and no account needed.
The tokens of Claude Code that one point of Claude’s usage gauge stands for, as the median of the accounts on a plan. It is shown for the 5-hour session and for the week.
This page cannot say yet. When a plan has enough accounts, it shows the measured ratio next to the promised one. A gap is a question, not proof.
Nothing here says so, and a measurement of this kind cannot prove it. Anthropic announces its changes, and our dated timeline lists them with their sources.
UsageMeter sees the tokens of Claude Code turns, through the UsageMeterLink app. Chats on claude.ai bring no tokens, so they cannot be measured this way.
A range: 5 to 9, 10 to 24, or 25 and more. Never the exact number, and nothing at all below 5.
No. Each account counts once in a median, with no name or account shown. The privacy policy describes what is kept.
Checked against Anthropic’s pages on October 30, 2026. Anthropic can change these rules at any time.
UsageMeter is an independent tool, not affiliated with Anthropic. Claude and Claude Code are Anthropic’s products.