SUMMARY Precambrian palaeointensity measurements provide fundamental constraints on the evolution of the deep Earth. Core evolution models predict trends in dipole moment on billion-year timescales that can be tested by palaeomagnetic records. Here, we report new palaeointensity results from the recently identified ∼2.62 Ga Yandinilling dyke swarm of the Yilgarn Craton, Western Australia, and consider them alongside published measurements spanning 500 Myr across the late Archaean to earliest Proterozoic. Rock magnetic and scanning electron microscopy analysis confirm that the magnetic mineralogy is fine-grained magnetite, appearing mostly as exsolved lamellae with ilmenite. Six sites produced acceptable palaeointensity estimates from thermal and microwave IZZI protocol Thellier experiments and from double-heating technique Shaw experiments. These site mean values of 9–26 µT translate to virtual dipole moments of 11–44 ZAm2 that are considerably lower than today's dipole moment of ∼80 ZAm2 and the value predicted for this time period by some thermal evolution models. Their average (median = 41 ZAm2) is, however, similar to the long-term average during both of the intervals 2300–2800 Ma (median = 44 ZAm2; N = 103) and 10–500 Ma (median 41 ZAm2; N = 997). While there is little evidence for a substantial net change in average dipole moment between the late Archaean and Phanerozoic, there is preliminary evidence that its variance has increased between the two intervals. This lower variance more than two billion years ago supports the idea that the geodynamo, even while not producing a stronger magnetic field, was more stable on average at the Archaean–Proterozoic transition than it is today.