Model techniques map
Techniquesoptimizationtraining parallelism

general family · filed under optimization

Expert Parallelism

A model-parallelism family that distributes experts across ranks; the evidence does not specify a more particular expert-placement method.

Also called MoE Expert Parallelism (EP), expert parallel, expert-parallel traffic.

sources
10
models
6
labs adopt it
5
strongest
used

How sources treat it

One count per evidence span, weakest treatment to strongest.

optional 2used 11

Documented in

Further reading

Picked by hand, not extracted: where to read more, not evidence for anything on this page.

Evidence

13 spans quoted from the sources, strongest treatment first.

--enable-expert-parallel

usedinference servingin GLM-5.3vLLM

keeping expert-parallel traffic inside one high-bandwidth domain

usedinference servingin Kimi K3Moonshot AI

128-way expert parallelism

usedsoftware implementationin Nemotron 3 UltraNVIDIA

wide EP the practical choice for high-throughput serving

usedinference servingin Nemotron 3 UltraNVIDIA

use --language-model-only to skip loading the vision encoder and free up memory for KV cache, and enable Expert Parallelism.

usedinference servingin vLLMvLLM

enable-expert-parallel

usedinference servingin Nemotron 3 UltraNVIDIA

We utilized a hybrid approach combining Attention Data Parallelism (DP) and MoE Expert Parallelism (EP)

usedinference servingin GLM-5Z.ai

Megatron’s Expert Parallelism (EP) groups can span multiple racks

usedsoftware implementationin Nemotron 3 UltraNVIDIA

expert parallelism (EP, which distributes full routed experts across ranks)

usedinference servingin Nemotron 3 UltraNVIDIA

vllm serve ... --enable-expert-parallel

usedinference servingin Step 3.7 FlashStepFun

sglang serve ... --ep 4

usedinference servingin Step 3.7 FlashStepFun

--enable-expert-parallel

optionalinference servingin Nemotron 3 UltraNVIDIA

--enable-expert-parallel

optionalinference servingin Step 3.7 FlashStepFun

Filed alongside

Other methods under optimization :: training parallelism.