Leap Seconds, 1972 to Now
Download all leap seconds since 1972 as CSV or JSON, and see the page rebuild each TAI-UTC offset and check it against two sources.
Loading the files...
Data: CC0 1.0 (LICENSE-data.txt). Code: MIT (LICENSE-code.txt). The two source files (A, B) keep their own terms (see the README). Made by Claude, an AI model.
What the columns mean
effective_utc_date- day the new offset starts (UTC)
mjd- Modified Julian Day of that date
tai_minus_utc_s- TAI minus UTC in whole seconds from that date
change_s- this offset minus the one before (empty on the starting row)
leap_second_at_utc- the extra second, 23:59:60 on the day before (empty on the starting row)
The first row sets the offset UTC started with on 1 January 1972; it is not a leap second. Each later row is one leap second, so the table has one row more than the number of leap seconds. Earlier years, when the UTC second was stretched, are not covered.
Rebuilt in code
Starting from the first row's offset and adding each row's change_s gives a running offset. The page then compares that running value with the tai_minus_utc_s written in each row, and checks that each change is a whole step, each date is the 1st of January or July, each MJD matches its date, and each leap second sits on the day before.
The chart is drawn from the rows loaded above; each step is one leap second.
Checked against two sources
Each row is compared with the same date and offset in two files kept in this folder, parsed in code on this page. The test below also reads the header text of each file for where it says its numbers come from.
The first button changes one offset in the page's copy of the rows (not in the files) so you can watch the checks fail; the second undoes it.
All rows
| Effective | Leap second at | TAI-UTC | Change | Rebuilt | Source A | Source B |
|---|
The 2022 CGPM decision
The 27th General Conference on Weights and Measures (CGPM) in 2022 passed a resolution on the future of UTC that is reported to ask for leap seconds to stop by 2035. Unverified: the BIPM resolution page (bipm.org/en/cgpm-2022/resolution-4) turned up in a search, but this page's builder could not open it, so the wording and the year 2035 come from search summaries only. What the 28th CGPM (October 2026) decided was not found; nothing here depends on it.