They have removed ability score requirements for multiclassing. This is the 13 in different stats that you need in order to multiclass. This change definitely makes multiclassing more accessible and could buff many builds, but few would say it is game breakingly overpowered. I tried to edit out previous mentions of ability score requirements.
I am not really sure what to make of this line, but it sounds like it is going to buff multiclassed casters:
The other thing we changed is how magic users use spell slots, making it less punishing to level up more than one magic class. One of the issues with multiclassing is that if you multiclass early in the game, you don't get strong abilities like "Fireball" at the same level as a "pure" class. But we wanted players to be able to multiclass from the beginning of the campaign, without necessarily having to wait for higher levels, so we had to tweak the resource usage a bit
Until we get clarification on what point 2 means, I am not going to account for it. It could significantly change anything with regards to multiclassing casters. I still think the below guidance can be immensely beneficial for understanding how multiclassing works, and then you can apply the modifications to your build based off the above rule changes.
Introduction and Assumptions
The D&D 5e rules for multiclassing can be found here for free. Larian may adjust these rules in BG3. Larian is already homebrewing several of the rules regarding race, class, and subclass options. Some D&D 5e builds may not work well in BG3. And BG3 may have some really good multiclassing builds that wouldn't be possible in tabletop. Until we know the exact details and to get people a basic understanding of the rules, I am going to assume that BG3 will be in line with D&D 5e except in areas we already know this will not be the case.
There are many posts out there with regards to D&D 5e multiclassing. This post will be unique in that it will account for the subclasses, races, and level cap which are found in BG3. But once again, we don't know all the details yet to plan a complete character builder, so I am treating any unknown features as though they work like in D&D 5e.
You can try to make your own D&D 5e multiclassed character using character creators like those on D&DBeyond. A free account I believe gets you 2 or 3 characters you can make at a time, and you will be very limited on what class, race, and subclass options you can pick. But you can dabble in multiclassing, pull up a character sheet for the character you made, and see how things work.
Edit to add this video by MortismalGaming which closely aligns with this post.
Opportunity Cost
It is difficult to "mess up" a D&D 5e character, but Multiclassing without a plan can be one way to mess up your character. Multiclassing does not come for free. Every level you take in some other class, means one less level in your main class. It is up to you to weigh whether this trade-off is worth it. For example, in the first of the below builds the character will finish BG3 at Barbarian 9/Fighter 3. Those three levels in fighter grant some great features, and many people would say it's worth the multiclass. But is it worth losing the subclass feature, feat/ability score increase, and relentless rage that a barbarian gets between levels 10-12 in their main class? That's for you to decide. If you are going to multiclass, then do it with a plan in mind and always weigh the opportunity cost.
It is worth emphasizing that multiclassing can result in you losing a feat/ASI. Most classes only get feats/ASIs every 4 levels (fighters or rogues can get additional ones at 6th or 10th class level respectively). Missing out on one or more of these can be extremely detrimental to your character at higher levels.
Common Multiclassing Confusion and Pitfalls
I am not going to cover all the multiclassing rules. They are described in depth here with examples. The basics are that when you level up, you can choose to increase the level in your existing class. Or you can take a level in a different class. So a second level character may be a Level 1 Fighter/Level 1 Wizard, getting most of the benefits of each class. But I will provide emphasis to some common sources of confusion and pitfalls.
What do you get from your second class? - You get almost everything that your second class has to offer. The biggest outliers are the starting proficiencies. You do not get any additional saving throw proficiencies by taking first level in a second class. You only get armor, skill, and weapon proficiencies from your second class that are shown in the table found in the rules. If your second class has a subclass which grants proficiencies at first level, then you DO get those proficiencies as well. The other big things that you do NOT get to stack from a multiclass are unarmored defense, extra attack, and uses of channel divinity. A 5th level paladin gets to make an extra attack when they take the attack action, and a 5th level barbarian gets to make an extra attack when they take the attack action. But a barbarian 5/paladin 5 does NOT get three attacks when they take the attack action (barring some external buff of some kind). Monk and barbarian unarmored defense do NOT stack, you get one or the other.
Class Level vs. Character Level vs. Spellcaster level - For the most part if you are trying to see what your character can or cannot do, you need to only look at the features granted by their level in each of their classes. The two main things that scale off Character Level are proficiency bonus and damage scaling from cantrips (0 level spells). There are other things that scale of character level, such as the Tough feat. But we are really getting into nuances here. Cantrip damage scaling and proficiency bonus are the big two. Spellcaster level is covered below.
Multiclassing Spellcasters (ignoring Warlocks) - Multiclassing spellcasters is the most complicated part. You determine your spells known/prepared (including cantrips) by looking at what spells you would know or prepare for each class with the Spellcasting feature. And you must use the spellcasting ability (Cha, Int, Wis) for the class which gave you that spell.
However to determine the number of spell SLOTS you must determine your spellcaster level, and then use the caster level table provided in the rules (which is identical to the spell slot table for any full caster). So while a second level sorc would only have three spell slots of 1st level, and a third level paladin would only have three spell slots of 1st level, a sorc 2/paladin 3 does NOT have six spell slots of 1st level. This character's spellcaster level would be third. 2 spellcaster levels from sorc, and only 1 spellcaster level from paladin since paladins are "half-casters" (like rangers), so you divide the class level by two and then round down to determine the spellcaster level. This means the character would have three first level spell slots, and a second level spell slot. But they won't actually know any second level spells, meaning that the second level spell slot can only be used to upcast the first level spells that they actually know.
While the basic rules version of multiclassing don't cover Eldritch Knight Fighter and Aracne Trickster Rogue, they follow the same rules as paladins and rangers except that EK fighter and AT rogues are 1/3 casters. Meaning you need to divide their level by 3 and then round down to get their effective spellcaster level.
Multiclassing Warlock with other casters - Note that warlocks do not have the Spellcasting feature. They have Pact Magic. Pact Magic and Spellcasting are separate. The warlock spell slots are in addition to whatever spell slots you have from your effective caster level as described above. When you take a short rest, you only get your warlock spell slots back (not counting something like Wizard's Arcane Recovery feature). However the warlock spells slots are interchangeable with spell slots used to cast spells from other classes, or class abilities like divine smite.
Spellcasting abilities (Cha, Int, Wis) - Didn't I already bring this up in italics in the "Spellcasters" section? Yes. But it is so important that I'm going into it with more depth. If somebody shows you a high elf character with levels in Sorc X/Druid Y, but don't see 2 reasons why this would possibly be a bad idea then keep reading. Any spells you learn from being a sorc use Cha for determining the chance to hit, or the target's ability to resist the effects. Similarly, any spells you learn from being a Druid will use Wis for determining the chance to hit, or the target's ability to resist the effects. Multiclassing between these two classes is possible, but you have to be careful about it. You either need to roll for stats (still pending verification that this will be an option in BG3) and get really good Cha and Wis scores to be able to use all Sorc and Druid spells effectively. Or you focus on one of these spellcasting stats (Wis or Cha). If you pick Wis, then you can choose whatever Druid spells you want, and just make sure you only pick Sorc spells that don't require dice rolls to hit or save, such as buffs like mirror image or jump. Vice versa if you pick Cha as your main casting stat, you will need to be careful about what Druid spells you pick. I picked high elf for this hypothetical character, because as part of picking this race they get a cantrip that scales off Int and a lot of people miss this detail, wondering why their sorc is using Int to cast ray of frost. Similarly, a drow's Faerie Fire scales off Cha. Always look at what the casting stat is for a racial spell, and if it requires an attack roll or saving throw.
Making a spellsword or "gish" character - BG3 does not include the Hexblade class, which is overtuned at first level and has become an overused staple of "gish" builds. This is because the hexblade subclass allows you to use Cha as both your spellcasting ability and your ability to determine accuracy and damage with weapon attacks. BG3 also does not seem to include the "SCAG" cantrips of booming blade or green flame blade. These spells help casters that don't have extra attack keep up in melee combat with martial characters that do have extra attack. It is a bit trickier to make gish characters without these options, but still doable. Single class options exist such as eldritch knight fighter, arcane trickster rogue, ranger, paladin, warlock (pact of the blade at level 3, thirsting blade at level 5), swords bard, valor bard. But if you want to make a multiclassed gish, then you will want to get to extra attack as soon as you can. Ideally level 5, though level 6 is doable. If you are at level 7 and playing in melee without extra attack or the SCAG cantrips, you're going to have a bad time.
Multiclassing martials - Multiclassing martials is pretty simple, once you understand that extra attack from different classes do NOT stack. But similar to above with gish characters, you want to get to extra attack as fast as possible. You can afford to perhaps take a dip outside of your main class. But you absolutely want to have extra attack by the time you are level 7 as a martial character.
Popular & Likely Viable BG3 Multiclass Builds
The below are common multiclass builds in D&D 5e which may work in BG3. I also added a couple builds that may work better in BG3 than they do in tabletop. These are by no means the only possible multiclass builds. There are tons and tons of them, and I myself am considering making my first character a paladin/rogue multiclass which I won't cover here since it isn't what folks would view as common or particularly synergistic.
Crit Fishing Half Orc - Barbarian 9/Champion Fighter 3
When a half-orc crits, they roll an extra damage die. When a 9th level barbarian crits, they roll an extra damage die. Champion fighters have double the crit chance (they crit on a 19 or 20 on the attack roll). And if a barbarian reckless attacks, you get advantage on the attack roll which further increases your odds of getting a crit (but also makes you a bit more vulnerable). If you want to go this route, make sure you use a two-handed weapon with a d12 damage die unless you find something extraordinary for your build. If you use something like a greatsword with a 2d6 damage die, then your extra damage dice from half-orc and barbarian would bring it up to 6d6 damage (average 21) on a crit. Whereas is you use a weapon like a battleaxe with a d12 damage die and crit then you will do 4d12 damage (average 26) on a crit. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Tanky Moon Druid - Moon Druid X/Barbarian Y
Moon druids can wildshape, and when they do they basically get a massive HP pool on top of their existing one. Barbarians can also rage, and while raged get only takes half damage from piercing, slashing, and bludgeoning damage (the three most common types). If Larian adds Totem (a.k.a. Wildheart) of the Bear, then this rage damage resistance expands to instead half ALL damage except psychic. If bear totem is in, then I think the best split is Druid 9/Barbarian 3. But you could argue for other splits as well. One common point of confusion is that people thing unarmored defense carries over into wildshape. The rulings on this are various, but technically you only get the "natural armor" of your wildshaped form, and do not get any bonus to your AC from your barbarian unarmored defense. Some DMs let players use unarmored defense in place of the wildshpaed form's natural armor if unarmored defense is higher, but note that your dex and con change to that of the creature you are wildshaped into. Under no circumstance should unarmored defense stack with wildshaped natural armor. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Tanky Caster version 1 - Fighter X/Squishy Caster Y
The "squishy caster" in this scenario is usually a Wizard. It also works with lore bards, warlocks, and sorcerers, but these classes will often do a similar build where they choose paladin instead of fighter as will be discussed below. By taking your first level in fighter you get proficiency in (among other things) medium armor, shields, and con save proficiency which is great for maintaining concentrations on spells. As well as a fighting style, and defense is a great option for this. Plus second wind for a bit of self-healing. By making your caster start as a fighter with 14 dex you'll have medium armor, a shield, and the defense fighting style for a total of 19 AC. Or you can start at 13 Str and go the heavy armor route with chain mail. Or you can go 15 Str and go the heavy armor route, and get 21 AC with plate mail, a a shield, and defense fighting style. This is enough to get from the fighter class for some people, but others will take a second level in fighter at some point (maybe character level 3) for Action Surge. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Tanky Caster version 2 - Cleric 1/Wizard 11
I guess this works on other squishy casters like sorc or bard or warlock, but I never hear people talk about doing this on anything but wizard. So I'll just specify wizard on this one. This build is taken for similar reasons as the above Tanky Caster version 1 build; provide armor class to a squishy caster. The downsides are that you don't get the defense fighting style, second wind, or the option to get action surge. What you DO get is another level in a full caster class, which fighter does not provide. This benefit may increase the number and level of spell slots you have through the early and middle stages of the game. But 11th level and 12th level caster have the same number of spell slots so it does not make a difference near the end of the game. You also get the spells known from a cleric. This includes great spells that don't require you to make any attack rolls or force saves, such as guidance or bless. So having the bare minimum 13 Wis is fine. And while healing word from the cleric does scale the healing slightly off of your Wis modifier, the amount healed is not really relevant when restoring an unconscious ally to life in the middle of combat. Lastly, you get the level 1 cleric subclass benefits. Many will pick a subclass that grants heavy armor proficiency such as life, nature, tempest, or war. I personally encourage knowledge domain to get expertise in two intelligence based skills that your wizard will already be great at. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Sorcadin - Paladin X/Sorcerer Y
This can be very similar to the above "Tanky Caster" where you get defensive features from Paladin, but are primarily a caster. But more commonly people play Sorcadin as a spellsword "gish" build. They get up close and smash things over the head, and they use all the extra spell slots they get from full caster levels in sorcerer to do divine smites and deal extra damage. Sorcerer is a popular pick because they get a lot of spell slots (as compared to warlock), and sorcery points to convert into more spell slots. Or they can use metamagic to cast powerful buff spells like haste as a bonus action. I personally feel that with a level 12 cap, it is ideal to start with Paladin 6 to get extra attack and the amazing paladin aura, then go the rest of the levels in sorc. Others may say to go 5 levels paladin and 7 in sorc to get access to 4th level sorc spells. Or may go 7 levels in paladin to get the improved paladin aura from your subclass, and 5 in sorc. You pick what you prefer. Larian gave white draconic bloodline sorc the spell Armor of Agathys as a spell known, so that is the sorc subclass I would pick with known info.
I had to edit this one because I forgot paladins get heavy armor proficiency only if that is your starting class. Caster focused Sorcadins who are mostly going sorc and only dipping into paladin for tankiness may want to opt for starting with 14 dex, starting as a sorc to get con save proficiency, then take your level(s) in paladin. Though starting as a paladin and putting 13 or 15 in Str and using heavy armor is also valid. But Str based melee Sorcadins should likely start in Paladin, unless you are doing a very niche and nuanced Dex sorcadin build. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Bardadin - Paladin X/Bard Y
This is very similar to the Sorcadin above, except bard instead of sorc. You can make a caster heavy focused baradin or a melee focused one, just like the sorcadin. However unlike Sorcadin, the bard has the valor and swords bard subclass. This means a melee focused Bardadin will focus on swords or valor bard for the first 6 levels to get extra attack and then 2 levels in paladin to unlock divine smite, and then the rest can go in bard. But starting as a bard and multiclassing into Paladin means you do not get heavy armor proficiency, so you'll want to be dex based if you go this route.
This build will do a similar amount of damage as the melee sorcadin since you don't have metamagic to make more spell slots, but you also have more spell slots available and at higher levels due to all your bard levels. Sorcadin probably wins on damage if you are using metamagic to quicken cast buff spells. But Paladin already has a ton of damage potential, and you are supplementing it with your bard spell slots so Bardadin damage is nothing to scoff at whatsoever. However you will be better at out-of-combat support through extra skill proficiencies, jack of all trades, and bardic inspiration than a sorcadin would be. I would play a bardadin over a sorcadin any day thanks to this utility. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Lockadin - Paladin 10/Warlock 2
This is very similar to the Sorcadin above, except warlock instead of sorc. You can make a caster heavy focused lockadin or a melee focused one, just like the sorcadin. The melee focused lockadin takes advantage of getting the warlock's spell slots that come back on a short rest on top of their paladin slots. So they smite away, and then get back two of their spell slots back on a short rest. This multiclass is a lot more common with the warlock's hexblade subclass. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Bardlock - Warlock 2/Bard 10
One of Bard's greatest down sides is a lack of a good damage source. But they are Charisma casters, warlocks are charisma casters, warlocks get eldritch blast which is a decent damage cantrip, but made into a great damage cantrip with the warlock's agonizing blast invocation at second level. So the build usually goes taking 2 levels in warlock, the rest in bard. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Sor-lock-bard-adain
Some combination of Paladin, Warlock, and the rest in either Bard or Sorc. Using the principles of above four builds. Maybe you make a paladin gish with the benefits of sorc and warlock spell slots. Maybe you make a bard-lock eldritch blast spammer, but dip into paladin for armor and shield proficiency. There are tons and tons of options between the 4 charisma casters. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The Ambush Bard - Bard X/Rogue Y
This one is actually really simple once you get the hang of it, but a bit complicated to explain without accompanying images. I think the most efficient way to explain this build is with this short video. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The Ambush Ranger - Assassin Rogue X/Gloomstalker Ranger Y/Fighter Z
It is worth noting that Larian has already made several changes to the ranger class. And outside of this build, almost nobody plays the rogue's assassin subclass in tabletop D&D 5e so it may also be adjusted at release. But per tabletop, this build has a devastating opening turn of combat. Especially if it opens from stealth.
Gloomstalker gets bonuses that make them very difficult to detect in darkness by most creatures, helping them catch their enemy by "surprise."
Gloomstalker ranger gets to add their Wisdom bonus to their initiative rolls, helping them go early.
Assassin rogue gets advantage on any creature that has not gone yet in combat.
If you started the combat from stealth or otherwise caught the enemy off guard, then any character that is caught unawares during the first round of combat has the "surprised" condition. And assassin rogues automatically crit when they hit a surprised target
Gloomstalker rangers get to make one extra attack on the first turn, and if that hits then it does an extra 1d8 damage.
If you are dual wielding or using the crossbow expert feat or polearm master feat then you can make an additional attack using your bonus action.
Action surge, and get three more attacks.
Put all this together and you get 2 attacks from a level 5 ranger or fighter's attack action, another attack doing an additional 1d8 damage from gloom stalker, and a bonus action attack if you are properly equipped with equipment and possibly feats for a total of four attacks. Now add at least 2 levels of fighter to get action surge, which gives you another 3 attacks (2 from extra attack, 1 from gloomstalker opening round which does an additional d8 of damage) attacks for a total of 7 attacks on this absolutely absurd opening turn. One of the above attacks will also do sneak attack damage. Any of these attacks made against targets behind you in initiative will have advantage, and if the target is surprised then each of these attacks automatically crit which doubles the amount of damage dice rolled. And with all these possible crits, half-orc is looking awfully tempting for the extra weapon damage die for each attack that crits.
Once you get past the first round of combat, this build significantly quiets down. But after such a devastating opening, the enemy is usually down a few members. The exact level split may be up in the air, pending possible changes to assassin rogue and the ranger class. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The "Great"berry Healer - Life Cleric X/Druid Y or Life Cleric X/Ranger Y
When a healing spell cast by a life cleric does healing, it does an additional amount of healing equal to your Wisdom modifier. The goodberry spell in tabletop summons 10 berries, each of which does 1 healing. And in tabletop the life cleric additonal healing does interact with each goodberry consumed. Meaning that a single cast of good berry will do somewhere between 40 to 60 hitpoints of healing, depending on a typical wisdom caster's Wisdom modifier. But this spell has been changed in BG3 Early Access at least to instead summon four berries, each of which does 1d4 healing (this may change on launch). On its own this change is a roughly equivalent amount of healing. But if BG3's goodberries interact with life cleric's additional healing bonus at all then this will significantly cut down the expected amount of healing to closer on the order of 22 to 30 hp. Edit: In BG3 Early Access this interaction will also only work if the life cleric who cast the goodberry spell is the one to consume the "great"berries, it does not work if another party member eats the "great"berries. It is still extremely powerful for a first level spell, but not as good as in tabletop. All you need is 1 level in life cleric, and the goodberry spell which can be obtained by 1 level in druid or 2 levels in ranger. You can make any of these three classes into your main class after that. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The Angry Rogue - Barbarian X/Rogue Y
For a rogue to get sneak attack they either need to have advantage on the attack roll, or the target must have another hostile (not counting the rogue) within 5 ft of it while the rogue does not have disadvantage. This is usually achievable, but not always. If only there was some way to just give yourself advantage... This is where two levels in barbarian comes in so you get Reckless Attack. You get advantage on your attack rolls, but enemies will have advantage on attack rolls against you for one round. Fortunately rage and uncanny dodge are there to help your defense a bit. The main way to build this would be to go Strength based, as sneak attack only requires that you use a weapon with the finesse property, but not necessarily that you use dex on your attack rolls. So you can sneak attack by wielding a rapier for example with strength. And while you are at it, apply bonus rage damage if you are raging. If you go this strength route you will probably want to take your first level in barbarian, because that is the only way this build is getting medium armor proficiency (unless you pick the mountain dwarf as your race, or the moderately armored feat). That is how most tabletop players do this build, and I'd recommend doing 2 barb and 10 rogue to catch the bonus ASI/Feat that rogues get at level 10.
But in BG3 it may be possible to do this with a dex based build. Normally in 5e you do not get the extra barbarian rage damage to your attacks unless you use strength, and you cannot recklessly attack unless you use strength for the attack roll. BG3 Early Access does not impose either of these restrictions, and a Dex based barbarian is a good bit tankier while unarmored than a strength based barbarian is while unarmored. Plus a Dex based barb can start as a rogue and get the extra skill proficiencies that come with this.
Edit: Furthermore in BG3 you may want to take barbarian to level 3 and pick berserker subclass, and pick the thief subclass for rogue. This will give you two powerful bonus action attacks you can take as a barbarian/rogue multiclass when raging if these subclasses launch the way that they are in Early Access. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The Classic Eldritch Knight Gish - Eldritch Knight Fighter X/Abjuration Wizard Y
The fighter is great because it makes lots of attacks and gets lots of feats. The eldritch knight subclass adds a bit of magic on there. But the magic you can use is severely limited by both spell selection and number of spell slots, since eldritch knight is a 1/3 caster. This is where wizard comes in to shore up the eldritch knight's short comings, by adding more spells known and more spell slots. Abjuration wizard is great, since you'll likely be casting the shield spell a good bit with this build and each time you do you'll get a bit of temp HP thanks to abjuration wizard's Arcane Ward feature. I personally would go EK fighter 7 and abjuration wizard 5. Or EK fighter 8 and abjuration 4 if you prefer an extra ASI/Feat over third level wizard spells. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
The Skill Monkey Support Bard - Sorc 1/Knowledge Cleric 1/Lore Bard 10
I meant to add one of these to the post, but am glad I did not because this commenter provided a better one then the one I was going to do.
Absurd Lightning Damage - Tempest Cleric 2/Storm Sorc 10
Storm Sorc is typically viewed as one of the weaker subclasses in 5e, and many are hoping to see it buffed in 5e. However Larian has a homebrewed rule in BG3 Early Access that causes lightning damage to do double damage against targets which are "wet." And when a storm sorc casts a leveled spell that does lightning or thunder damage, then you can cause all enemies within 10 feet of you to take lightning or thunder damage equal to half of your Sorc level (cantrips don't trigger this effect). So if an enemy is wet (possible via a quicken cast of create water, an ally setting it up by throwing a water flask or other means, etc.) and you cast a lightning or thunder leveled spell like thunderwave or chromatic orb, you will do double damage with the lightning spell and damage to all nearby enemies equal to your sorc level. This is potentially very powerful without multiclassing.
But wait, there's more. Tempest domain clerics get channel divinity at second level allowing them to roll max damage on all the damage on a lightning or thunder spell once a short rest (essentially double damage). So by combining this with storm sorc and the BG3 EA wet condition making enemies vulnerable to lightning damage, this means that once a short rest you can do 4x damage with a lightning spell on a wet enemy, and simultaneously deal damage equal to your sorc level on all enemies near you. And you can use metamagic to do even more damage by enhancing it or twin casting the spell if action economy allows. This is a gimmicky but viable build in tabletop, and if Larian's homebrewed "wet = double lightning damage" stays in then it will be even more gimmicky but also more viable in BG3. I recommend first level sorc to get con save proficiency, then two cleric, then the rest sorc. You do not need to have a high Wis for this build, Cha is your main casting stat. Just make sure that the cleric spells you pick are ones that don't require an attack roll or save. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
Sorlock - Warlock 2/Sorcerer 10
Note it is almost certain that the once mighty tabletop build referred to as the "coffeelock" where you get infinite spell slots will NOT work in BG3. But even ignoring this rulebending build, Sorlocks are still viable. Typically the premise is that sorcs can do great burst damage. Wouldn't it be nice if they can do some extra consistent damage on top of that? So start sorc, take a two level dip in warlock for eldritch blast and agonizing blast, then go sorc the rest of the way. This also gives you some extra warlock pact magic spell slots which come back on a short rest so you can convert into sorcery points to fuel your metamagic shennanigans. Cast a powerful eldritch blast with your action, and follow it up with a quickened fireball with your bonus action or quickened haste to buff your paladin who needs to go charging into combat. Assuming relevant rules are implemented per D&D 5e, and you should weigh the opportunity cost.
A note on Monk
The best monk features often rely on Ki points, which you can only get more of by taking more levels in Monk. If you want to multiclass out of monk, most find that the opportunity costs are too high and you would have been better off just staying with monk. Taking a dip into monk grants you melee combat focused features that only work if you are unarmored, so will be useless on most characters. The monk's unarmored defense does not stack or otherwise interact with unarmored defense from a wildshaped druid, or a barbarian's unarmored defense. These reasons are some of the more prominent ones for why monk multiclasses are not common in tabletop. Larian has said they are making changes to monks in BG3. Depending on these changes, monk may become a more tempting multiclass target
I'll link to some comments with monk multiclass ideas from commenters here:
Spore druid/monk
Shadow monnk/battlemaster fighter
Shadow monk/thief rogue
A note on Thief Rogue
The thief rogue's Early Access ability to get two bonus actions may be the most power shifting feature in BG3. Many martial casters will get benefit from dual wielding, crossbow expert, or polearm master with a 3 level dip in thief rogue. This is especially the case for ranger, which is notoriously starved for bonus actions. Bonus action controllable spells like spiritual weapon or flaming sphere just got their damage doubled by this. Currently in BG3 early access, Larian does not restrict you from casting a leveled spell with your action and bonus action. This means that a Thief Rogue 3/Fighter 2/Sorc 7 would be able to cast a spell with their action, action surge and get another action to cast a leveled spell, use metamagic to cast a powerful spell as a bonus action, and use metamagic to cast another powerful spell with your extra bonus action. You will burn everything you have to do this, and you will be limited to 4th level spells, so maybe that will balance things. A raging berserker barbarian now gets two very powerful bonus action attacks when they rage, which could go great with the Angry Rogue build. But if thief rogue is going to launch as it is in early access, you should always ask yourself, "Is 3 levels in thief rogue worth the opprotunity cost.