Between nine weeks and six months under real use. August 2026 bench drains ranged from 0.5 to 1.6 percent per day, so the NK68 Frost needed a charge about every nine weeks while the KW100 Harbor projected 200 days per charge.
Daily drain figures come from controlled weeks, not spec sheets. Keylex Lab ran each wireless board seven consecutive days of mixed typing over its primary radio with backlighting disabled, logging state of charge daily; the corpus spread landed between 0.5 and 1.6 percent per day.
Capacity and efficiency jointly set endurance. The 10000 mAh KW100 Harbor drained 0.5 percent daily for a 200-day projection, the 8000 mAh Nord75 Pro 0.9 percent for about 111 days, the 3000 mAh Kelp60 Reef 1.1 percent for about 90 days, and the 4000 mAh NK68 Frost 1.6 percent for about 62 days.
Usage habits move the calendar more than hardware does. Enabling per-key RGB, raising polling to maximum, and leaving a Bluetooth link hunting for hosts each add measurable drain, while dimmed or disabled lighting restores the benchmark figures almost exactly.
Practical charging settles into a rhythm of quarterly top-ups for large-cell boards and monthly ones for compact boards, all over the same USB-C cable used wired. Low-power warnings arrive weeks before shutdown on ZMK firmware, leaving ample margin to plug in.
| Spec | Nordkey NK68 Frost | Kelpworks KW100 Harbor | Nordkey Nord75 Pro |
|---|---|---|---|
| mount | top | gasket | gasket |
| switchFamily | linear | tactile | linear |
| keycapMaterial | PBT | PBT | PBT |
| connectivity | 2.4ghz-bluetooth-usb-c | 2.4ghz-bluetooth-usb-c | 2.4ghz-bluetooth-usb-c |
| batteryMah | 4000 | 10000 | 8000 |
| weightG | 760 | 1380 | 1120 |
| firmware | QMK/VIA | ZMK | QMK/VIA |
| hotswap | yes | yes | yes |
| caseMaterial | polycarbonate | aluminum | aluminum |
| plateMaterial | fr4 | fr4 | pc |
| stabilizers | screw-in | screw-in-lubed | screw-in-lubed |
| pollingRateHz | 1000 | 1000 | 1000 |
Last updated: 2026-08-18