For a super ignorant person:
Both Llama 4 Scout and Llama 4 Maverick use a Mixture-of-Experts (MoE) design with 17B active parameters each
Those experts are LLM trained on specific tasks or what?
Both Llama 4 Scout and Llama 4 Maverick use a Mixture-of-Experts (MoE) design with 17B active parameters each
Those experts are LLM trained on specific tasks or what?
This generally works well, although there are lots and lots of caveats. But it is (mostly) a free lunch, or at least a discounted lunch. I haven’t seen a ton of analysis on what different experts end up doing, but I believe it’s widely agreed that they tend to specialize. Those specializations (especially if you have a small number of experts) may be pretty esoteric / dense in their own right.
Anthropic’s interpretability team would be the ones to give a really high quality look, but I don’t think any of Anthropic’s current models are MoE.
Anecdotally, I feel MoE models sometimes exhibit slightly less “deep” thinking, but I might just be biased towards more weights. And they are undeniably faster and better per second of clock time, GPU time, memory or bandwidth usage — on all of these - than dense models with similar training regimes.
So the net result is the same: sets of parameters in the model are specialized and selected for certain inputs. It's just a done a bit deeper in the model than one may assume.
I think where MoE is misleading is that the experts aren't what we would call "experts" in the normal world but rather they are experts for a specific token. that concept feels difficult to grasp.
that was AFAIK (not an expert! lol) the traditional approach
but judging by the chart on LLaMa4 blog post, now they're interleaving MoE models and dense Attention layers; so I guess this means that even a single token could be routed through different experts at every single MoE layer!
It's more of a performance optimization than anything else, improving memory liquidity. Except it's not an optimization for running the model locally (where you only run a single query at a time, and it would be nice to keep the weights on the disk until they are relevant).
It's a performance optimization for large deployments with thousands of GPUs answering tens of thousands of queries per second. They put thousands of queries into a single batch and run them in parallel. After each layer, the queries are re-routed to the GPU holding the correct subset of weights. Individual queries will bounce across dozens of GPUs per token, distributing load.
Even though the name "expert" implies they should experts in a given topic, it's really not true. During training, they optimize for making the load distribute evenly, nothing else.
While current MoE implementations are tuned for load-balancing over large pools of GPUs, there is nothing stopping you tuning them to only switch expert once or twice per token, and ideally keep the same weights across multiple tokens.
Well, nothing stopping you, but there is the question of if it will actually produce a worthwhile model.
so you mean a "load balancer" for neural nets … well, why don't they call it that then?
Even in the single GPU case, this still saves compute over the non-MoE case.
I believe it's also possible to split experts across regions of heterogeneous memory, in which case this task really would be something like load balancing (but still based on "expertise", not instantaneous expert availability, so "router" still seems more correct in that regard.)
They don't really "bounce around" though do they (during inference)? That implies the token could bounce back from eg. layer 4 -> layer 3 -> back to layer 4.
So if the model has 16 transformer layers to go through on a forward pass, and each layer, it gets to pick between 16 different choices, that's like 16^16 possible expert combinations!
Meta calls these individually smaller/weaker models "experts" but I've also heard them referred to as "bozos", because each is not particularly good at anything and it's only together that they are useful. Also bozos has better alliteration with boosting and bagging, two terms that are commonly used in ensemble learning.
Makes sense to compare apples with apples. Same compute amount, right? Or you are giving less time to MoE model and then feel like it underperforms. Shouldn't be surprising...
> These experts are say 1/10 to 1/100 of your model size if it were a dense model
Just to be correct, each layer (attention + fully connected) has it's own router and experts. There are usually 30++ layers. It can't be 1/10 per expert as there are literally hundreds of them.
The models get trained largely the same way as non-MoE models, except with specific parts of the model silo'd apart past a certain layer. The shared part of the model, prior to the splitting, is the "router". The router learns how to route as an AI would, so it's basically a black-box in terms of whatever internal structure emerges from this.