To supplement the other answers:
Most OpAmps (OA) have push pull “Totem Pole” drivers configured as complementary Darlingtons or CMOS drivers that can supply current to the load.
Push=Source using upper driver and
Pull=sink using lower driver.
Thru a series of differential gain amplifiers the voltage is amplified from 1e5 to 1e7 while reducing the input current by a similar order of magnitude.
THe output impedance in BJT’s may be around 200 Ohms open loop and then is reduced by the amount of gain leftover from the forward feedback gain. Or in other words the feedback gain reduces the Zout by reducing the differential input to near zero V while reducing the output impedance to near 0 Ohms for small signals.
For large signals the internal current limiting sensed by a shunt resistor cuts off the output stage to limit the output current then the voltage is said to saturated, the gain drops to zero and the current is limited to some level by design of OA. So you see , for small signals of current, it is pretty close to an ideal voltage source with 200 Ohms / 1e6 for a unity gain buffer being 200 uohms.
Adding additional complementary emitter followers to the OA output can increase the current gain by another factor of 100 or so and then becomes limited by temperature rise of these and drivers. These also push and pull the output voltage to match the input voltage of any OA differential or single ended input. Putting feedback at these connected emitters eliminates the crossover distortion. ( but reduces full swing by diode drops)
There are tradeoffs with Gain bandwidth (GBW) such that the gain drops to 1 at f=GBW so the impedance rises with f from near 0 to 200 Ohms at say 1MHz for GBW=1e6 and unity gain.