I am developing my own firmware, based on STM32CubeWL Firmware Package. I already optimised it further to 1.3uA by switching off bunch of GPIO’s I do not really need.
AT slave firmware will never be able to get there because UART is always on. STM32WL has also low power UART (LPUART), I do not know how much it consumes. I assume more than 2uA. Baud rate is limited to 9600.
It is also possible to make custom AT-slave firmware with a wake-up GPIO pin: when it is low, it goes into STOP2 mode where it will consume like 2uA and when high, it wakes up and ready to communicate via UART.