Certainly for the 6502, WozMon fitted into 256 bytes, but had access to the Apple 1 character printing and keyscanning routines. Once you take these dependencies into account the total will be somewhat higher.
Some form of VHF modulated or composite video output might be possible, as has been demonstrated on an ATtiny85. In this case it is written in C and has a full character set - so some code reductions might be possible.
Compared to the classic micros of the mid-1970s (6502, Z80) etc - we now have the advantage of speed.
An AVR can be overclocked to about 30MHz and most of it’s instructions are only 1 clock cycle long. Potentially this means that a 30 fold speed improvement over the fastest of the Z80 instructions.
If you remember that the Sinclair ZX80 started life with a 4K BASIC ROM. This contained an integer BASIC interpreter, monochrome 64 character tables, keyscanning and cassette interface.
Li-Chen Wang’s Palo Alto Tiny BASIC fits into as little as 1.77k bytes and was written for the 8080. As it used a serial terminal for communication - there was no need for keyscanning, character tables or video generation code.