Skip to content
KernelIndex
Search⌘K

MoE sparse routing and dispatch

31 eligible runs
moe

Mixture-of-Experts sparse routing and expert dispatch system. Routes each token to top-8 experts out of 128 total experts using gating network, performs expert computation with dynamic batching, and aggregates results weighted by routing probabilities.

Fastest reported
218.3µsmean · 95% of SOL
GeometricLicense unknown · unknown

Reported evidence · last observed 2026-08-20. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhidden_size = 2048 · num_experts = 128 · intermediate_size = 768 · num_experts_per_tok = 816 cases

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 16 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.1 · mean · 22 results · last observed 2026-08-20Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
218.3µs95% SOL
1.00×
Reported · license unknown · no source
2026-08-20

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
2
225.3µs94% SOL
1.03×
Reported · license unknown · no source
2026-08-19

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
3
233.1µs94% SOL
1.07×
Reported · license unknown · no source
2026-08-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
4
234.9µs93% SOL
1.08×
Reported · license unknown · no source
2026-08-15

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
5
310.7µs84% SOL
1.42×
Reported · license unknown · no source
2026-08-05

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
6
326.9µs83% SOL
1.50×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
7
327.9µs83% SOL
1.50×
Reported · license unknown · no source
2026-08-11

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
8
332.0µs82% SOL
1.52×
Reported · license unknown · no source
2026-07-23

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
9
361.9µs80% SOL
1.66×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
10
408.3µs77% SOL
1.87×
Reported · license unknown · no source
2026-07-17

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
11
438.2µs74% SOL
2.01×
Reported · license unknown · no source
2026-07-27

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
12
446.2µs74% SOL
2.04×
Reported · license unknown · no source
2026-07-16

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
13
456.2µs73% SOL
2.09×
Reported · license unknown · no source
2026-07-24

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
14
464.3µs72% SOL
2.13×
Reported · license unknown · no source
2026-08-08

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
15
479.7µs71% SOL
2.20×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
16
524.9µs68% SOL
2.40×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
17
543.6µs67% SOL
2.49×
Reported · license unknown · no source
2026-08-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
18
853.8µs52% SOL
3.91×
Reported · license unknown · no source
2026-07-24

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
19
858.2µs52% SOL
3.93×
Reported · license unknown · no source
2026-07-24

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
20
1.02ms47% SOL
4.68×
Reported · license unknown · no source
2026-07-10

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
21
1.65ms33% SOL
7.57×
Reported · license unknown · no source
2026-07-22

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →
22
5.75ms12% SOL
26.33×
Reported · license unknown · no source
2026-08-07

Measured exactly what you asked. Reported by source; not independently reproduced.

no sourcelicense unknownno install recipeRun detail →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
218.3µs
1.00×
Reported
license unknown · no source
unknown
325.6µs
1.49×
Reported
license unknown · no source
unknown
332.0µs
1.52×
Reported
license unknown · no source
unknown
357.6µs
1.64×
Reported
license unknown · no source
unknown
408.3µs
1.87×
Reported
license unknown · no source
unknown
438.2µs
2.01×
Reported
license unknown · no source
unknown
962.2µs
4.41×
Reported
license unknown · no source
unknown
1.44ms
6.59×
Reported
license unknown · no source
Show all 31 implementations ›
unknown
225.3µs
1.03×
Reported
license unknown · no source
unknown
233.1µs
1.07×
Reported
license unknown · no source
unknown
326.9µs
1.50×
Reported
license unknown · no source
unknown
361.9µs
1.66×
Reported
license unknown · no source
unknown
446.2µs
2.04×
Reported
license unknown · no source
unknown
456.2µs
2.09×
Reported
license unknown · no source
unknown
464.3µs
2.13×
Reported
license unknown · no source
unknown
467.9µs
2.14×
Reported
license unknown · no source
unknown
479.7µs
2.20×
Reported
license unknown · no source
unknown
517.3µs
2.37×
Reported
license unknown · no source
unknown
524.9µs
2.40×
Reported
license unknown · no source
unknown
543.6µs
2.49×
Reported
license unknown · no source
unknown
650.8µs
2.98×
Reported
license unknown · no source
unknown
853.8µs
3.91×
Reported
license unknown · no source
unknown
858.2µs
3.93×
Reported
license unknown · no source
unknown
1.02ms
4.68×
Reported
license unknown · no source
unknown
1.44ms
6.60×
Reported
license unknown · no source
unknown
1.65ms
7.57×
Reported
license unknown · no source
unknown
5.75ms
26.33×
Reported
license unknown · no source
unknown
11.8ms
53.95×
Reported
license unknown · no source
unknown
234.9µs
1.08×
Reported
license unknown · no source
unknown
310.7µs
1.42×
Reported
license unknown · no source
unknown
327.9µs
1.50×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesbf16 [batch_size, seq_len, hidden_size]
gate_weightbf16 [num_experts, hidden_size]
expert_gate_projbf16 [num_experts, intermediate_size, hidden_size]
expert_up_projbf16 [num_experts, intermediate_size, hidden_size]
expert_down_projbf16 [num_experts, hidden_size, intermediate_size]
norm_topk_probbool scalar
outputbf16 [batch_size, seq_len, hidden_size]
Axes and behavior
seq_lenvariable
batch_sizevariable
hidden_sizeconstant = 2048
num_expertsconstant = 128
intermediate_sizeconstant = 768
num_experts_per_tokconstant = 8
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias008_moe_sparse_routing_and_dispatchmodelqwen3-30b-a3b-instruct-2507sha256b4584cfdc5fa…
Sources: NVIDIA SOL-ExecBench (2026-08-20)last observed 2026-08-20How records are decidedJSON