The 28th CGPM, October 2026: Leap Seconds and a Redefined Second
Every four years the CGPM decides what the world's units mean. In October 2026 the agenda includes the death of the leap second and the clock that will replace caesium.
15 September 2026 · 7 min read · Science
From 13 to 15 October 2026, delegates from the member states of the Metre Convention meet at Versailles for the 28th General Conference on Weights and Measures, the CGPM. It is the body that decides what a kilogram, a second and a metre actually mean, and it meets roughly once every four years.
Two items on its agenda affect anyone who works with time.
1. The end of the leap second
Coordinated Universal Time (UTC) is kept within 0.9 seconds of the Earth's rotation by occasionally inserting a leap second. The Earth is not a good clock: it speeds up and slows down slightly, so 27 leap seconds have been added since 1972.
For computing, those extra seconds are a hazard. A minute that contains 61 seconds has taken down airlines, trading platforms and large parts of the web, because a great deal of software assumes minutes are exactly 60 seconds long.
At the 27th CGPM in November 2022, member states resolved to stop adding leap seconds by or before 2035, and asked for a concrete plan to be brought to the 2026 meeting. That plan is what is on the table now: how far UTC should be allowed to drift from the Earth's rotation before any correction, and who decides when a correction happens.
The practical effect for the rest of this decade is that clock time carries on as normal. The change is about what happens after 2035, when civil time is expected to run as an unbroken count of seconds.
2. A new definition of the second
Since 1967 the second has been defined by the caesium atom: 9,192,631,770 oscillations of the radiation from a specific transition in caesium-133. Every atomic clock, GPS satellite and network time server ultimately traces back to that number.
Optical clocks now do better. They use transitions at optical rather than microwave frequencies, roughly 100,000 times faster, which makes them more than an order of magnitude more accurate and more stable than the best caesium fountains. The leading candidates use strontium, ytterbium, aluminium and mercury ions.
The obstacles have never been about accuracy alone. A redefinition needs:
- Several independent clock types agreeing with each other.
- Clocks that run continuously rather than for a few hours at a time.
- International comparisons, including the recent campaigns that linked optical clocks in different countries over fibre networks.
The consultative committee on time and frequency set out a roadmap in which a redefinition becomes possible around 2030. The 2026 meeting is where the consensus gets tested: delegates can propose a specific option, keep the caesium definition for now, or decline to choose yet.
Why redefinitions do not change your measurements
This is the part that surprises people. When the kilogram was redefined in 2019 in terms of the Planck constant, no kitchen scale changed its reading. That is the entire point of a good redefinition: the new definition is engineered to hand back the same value as the old one, only with a better way of realising it.
A redefined second will still be the second you count with. What changes is the precision available at the top of the chain, which eventually filters down to satellite navigation, telecommunications, and experiments testing whether the constants of physics drift.
The pattern of recent decisions
| Year | Meeting | What changed |
|---|---|---|
| 2018 | 26th CGPM | Kilogram, ampere, kelvin and mole redefined by constants, in force from May 2019 |
| 2022 | 27th CGPM | New prefixes ronna, quetta, ronto and quecto; resolution to end the leap second by 2035 |
| 2026 | 28th CGPM | Leap-second plan; the roadmap towards an optical second |
Those 2022 prefixes are worth knowing if you work with data: ronna is 10²⁷ and quetta is 10³⁰, added because global data volumes were closing in on the top of the old scale at yotta (10²⁴).
What to watch for
If you maintain systems that handle time, the useful question is not "when is the second redefined" but "what does my stack do when leap seconds stop". Smearing schemes, leap-second tables and hard-coded offsets all carry assumptions that will expire.
Related reading
This article is for general information. Please verify figures against the relevant standard before using them in critical work. About accuracy