> LM394 spice model
Bob hates SPICE, so National only has SPICE models for the devices they adopted from other vendors like ComLinear.
What do you need to know that can't be adequately modeled by 2N2222 or 2N2906 type generic small NPN models? One "flaw" in SPICE is that any two transistors (same type and bias) match EXACTLY, so a super-match LM194 may not look any different than a generic transistor. LM194 has lower base resistance, but most SPICE models have improbably low base resistances (10Ω???). Most generic small NPN models give optimistic Beta values, not as high as an actual 194 but in that range. If you are relying on the 194's very high Beta, you can guestimate the conditions with pencil and paper. If you are relying on the 194's very high Beta, you may be in for trouble unless you account for variation of Beta and keep a very clean board.
Personally, I SPICE with 2N2222. If that isn't good enough, then I need a better plan, not a magic transistor. Then when I build it, I know that some 2N2222 will be more leaky and noisy than SPICE models say, so in critical stages I'll use a better device (or just try 2N2222/3906/etc: very often they work close-enough to theory).