There are ways to do what you want, although none are perfect.
High Toughness and Multiple Actions per Turn
This is the mathematically easy solution: If a boss has a Robustness 4 to 6 points above the norm, it will take a lot of tries to score a hit capable of taking them down. By at the same time giving them many things they can do per turn, you can keep an entire team "busy" with attacks/effects/secondary whatnots. This is how a dragon is statted out in the core rules.
However, this in my experience and opinion a bad option, because it just increases the incredible swinginess of Savage Worlds combat. No matter how high the Robustness, there is a nonzero chance the combat is over in the first turn, and on the reverse, there is a nonzero chance it will take "forever", or at least longer than the player characters can last.
Since wounding the boss is essentially random, the players won't feel very engaged during the battle. Their decisions will not matter much, it's all just waiting for that triple explosion on the damage dice, while hoping at the same time the BBEB won't get it first.
Similarly, multiple actions from a single source are more swingy than the same amount of actions from multiple sources. If the boss is Shaken, all actions are lost at once. And especially for multiple melee attacks, there is a severe focus fire problem, in which one guy gets hit by everything and thus immediately dies. Or in reverse, in case your players stay wisely at range, suddenly all those nifty multi-options are useless.
Some of the example monsters in the bestiary use this approach, in that they can only be damaged/wounded by certain specific things. E.g. a Vampire can only be Shaken by any result that's not sunlight, holy or a stake through the heart. This obviously makes the monster undefeatable until the weakness is found and exploited.
This is, while a bit better than just high Toughness, also a bad option my opinion. This is for two reasons:
Most players and GMs let their controlled characters fight until one side is dead. If however a monster is undefeatable by normal means, then this makes a combat encounter with that monster a foregone conclusion: Total Party Kill. As a GM, you need to re-think combat to have other end states than dying, and to make really sure that your players realize they have the option to try and disengage. This is not easy.
This kind of approach is inherently rail-roady: What the player characters do is meaningless until they find and exploit the weakness. Especially if the scenario to discover that weakness is badly designed, this will kill all enjoyment for the players as well.
Smart Bennie Use
Each GM Wildcard has 2 Bennies per Session, and in addition, the GM has 1 Bennie per Player around the table, which can be used at any time, even for Extras.
One of the main uses of Bennies are Soak rolls, negating a wound. This means, more than the Wound track, it's the amount of Bennies that is equivalent to the HP amount of other games. Once a boss can't Soak any more, it's usually over quickly, because the penalties of wounds make everything more difficult, and easier or their opponents.
Therefore, if you save the generic pool Bennies for the final confrontation, the Boss gets proportionally harder. With four players, he goes from Soaking 2 times to 6 times, so the combat will also last 3 times as long (on average).
Savage Worlds is a narrative system insofar that it just supplies mechanics and then leaves the interpretation of those mechanics in the hands of the GM. What does it look like to be Shaken? What real-world action, if any, is represented by a Soak roll (or Bennie spend in general)? This can be used to make a foe appear more dangerous than their raw stat block might suggest, by simply giving an unusual narrative explanation to the Soak/Recovery rolls of the character.
For example, one of the more memorable fights in my recent campaign was against a guy with rather average combat stats. If the main fighter of the group had been there, it would have been over quickly, as basically the only thing out of the ordinary was that the enemy ignored all wound modifiers.
However, the players involved became really desperate during the fight, because I narrated that guy as basically the Implacable Man. When the rogue opened the fight by a called sword attack to the head, I described the successful Soak roll as (instead of as usual, the wound being slight despite initial appearance) him continuing to function despite having a gaping head wound and no eyes left to see with. Later, he Soaked a wound acquired from the ranger charging on an inpromtu mount and impaling him with a lance. The description was that the man grabbed the lance and pulled it out by pushing rider and mount backwards over the ground. None of that had any mechanical impact, but it made the fight extremely memorable and impressive.
As desiging a combat encounter to be meaningful but difficult in Savage Worlds is not easy, a better option is to not rely on combat, which also meshes better with the genres Savage Worlds tried to emulate. What makes Toht and Belloq so problematic isn't that they can beat Indiana Jones in a fist fight, it's that they have a resource advantage with their Nazi backing and a greater willingness for amoral actions.
Therefore, instead of trying to shoe-horn a boss fight the players are expected to loose in there, create secondary conditions which make just attacking or killing the villain difficult or undesirable. This can be as easy as not having them be in direct contact with the heroes until the last act. A villain can be intimidating and difficult to overcome without being any kind of personal danger.
For example, in my current campaign, my group is currently facing a mage who they managed to easily defeat before. This time, they're a lot more careful, not because their opponent has better stats, but because the surroundings are different: Last time, it was an isolated encounter in a place that didn't care much. This time, he's a boss of a cult who has infilftrated the local law enforcement, and living in a place where any entanglement with the law on the part of the heroes might result in a civil war breaking out. Therefore, even if they were to meet that mage in the street, they couldn't deal with him as easily as last time.
While of course not really answering the question per se, the best option is to simply accept that Savage Worlds doesn't do one-on-one fights and duels very well. The system is designed for larger scale battles, with multiple participants on both sides.
Therefore, instead of trying to make that Big Beefy Boss work, it's best to instead supply a multitude of opponents, with differing abilities, who work together, to give a hard challenge for the team. This also prevents the anticlimax of the Boss going down in a single hit because of a lucky four-fold explosion.
This applies as much out of combat as in combat. By letting go of the idea of one central keystone villain, the narrative breadth and responsiveness of your scenario to the player's actions increases dramatically. I'll point at Justin Alexander's nice Article "The Principles of RPG Villainy" for a longer exploration of that subject.
And, because it should be mentioned because it's from an official source:
In the "Daring Tales of Adventure" booklet, the authors introduce a special rule to use for more Pulp-like scenario flow. Paraphrased because I've only got the German translation:
A GM Wildcard may at any moment in combat spend a Benny to Flee, that is, move their max speed and ignore all obstacles or attacks of opportunity preventing them from making that escape. They automatically suceed in all checks they'd need to make and can take that action even when it's not their turn during combat.
This allows specifically for recurrring villains, and prevents them from dying prematurely. But I'd also call it a bad option, because the notion on a boss dying prematurely and then just inventing a rule to prevent that are both symptoms of railroading.