a SPICE model
When you buy a real transistor or op-amp, the manufacturer often gives away a small text file, its SPICE model, that captures that exact part's measured behaviour. Dropping it into your simulation is like swapping a generic stand-in for the real actor: now the simulator behaves like the specific part number you intend to solder down.
A SPICE model usually comes as a '.model' line (for a single primitive like a diode or transistor, listing its measured parameters) or, for anything complex, as a subcircuit. You point your component at it by name. For example, a manufacturer's '.model 1N4148 D(Is=2.5n Rs=0.6 ...)' line fills the built-in diode equations with numbers measured from the real 1N4148, so its forward drop, capacitance, and recovery match the datasheet rather than a generic default. You download these from the part's product page and include them in your netlist.
Using the manufacturer's model is the difference between a rough sketch and a trustworthy prediction, especially for op-amps, where the model encodes the real gain-bandwidth, slew rate, offset, and supply behaviour. But two honest cautions: models vary in quality (some vendors model only typical behaviour, not the worst case, and some omit temperature effects), and a model is still just a model, it cannot include your layout, so confirm critical designs on the bench.
Swap a generic op-amp for a manufacturer's TL081 SPICE model and a transient run suddenly shows the real slew-rate limit, the output can no longer rise faster than about 13 volts per microsecond, a flaw the ideal model hid completely.
A manufacturer-supplied file that makes the simulation behave like one specific, real part number.
Vendor models vary in fidelity, many capture only typical behaviour and skip temperature or worst-case extremes. A good model sharpens a prediction but still cannot see your wiring and layout.