Skip to content
KernelIndex
Search⌘K

MoE sparse routing and dispatch

30 eligible runs
moe

Mixture of Experts sparse routing with top-2 expert selection from 64 experts, correction bias, weight normalization, token dispatch to expert MLPs with SiLU activation, and result aggregation. Includes shared expert computation.

Fastest reported
5.74msmean · 81% of SOL
Amir M. Mir | SF TensorLicense unknown · unknown

Reported evidence · last observed 2026-08-16. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 9 cases · mean latencytop_k = 8 · hidden_size = 8192 · num_experts = 64 · intermediate_size = 3584 · shared_intermediate_size = 71689 cases

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 9 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.1 · mean · 17 results · last observed 2026-08-16Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
5.74ms81% SOL
1.00×
Reported · license unknown · no source
2026-08-16

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

no sourcelicense unknownno install recipeRun detail →
2
5.83ms80% SOL
1.02×
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 →
3
5.93ms80% SOL
1.03×
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 →
4
6.09ms78% SOL
1.06×
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 →
5
6.10ms78% SOL
1.06×
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 →
6
6.11ms78% SOL
1.06×
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 →
7
6.18ms77% SOL
1.08×
Reported · license unknown · no source
2026-08-16

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

no sourcelicense unknownno install recipeRun detail →
8
6.19ms77% SOL
1.08×
Reported · license unknown · no source
2026-07-30

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

no sourcelicense unknownno install recipeRun detail →
9
6.20ms77% SOL
1.08×
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
6.32ms76% SOL
1.10×
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 →
11
6.39ms75% SOL
1.11×
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 →
12
7.58ms66% SOL
1.32×
Reported · license unknown · no source
2026-07-11

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

no sourcelicense unknownno install recipeRun detail →
13
11.5ms47% SOL
2.01×
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 →
14
11.8ms46% SOL
2.06×
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 →
15
15.7ms36% SOL
2.74×
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 →
16
16.3ms35% SOL
2.83×
Reported · license unknown · no source
2026-07-28

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

no sourcelicense unknownno install recipeRun detail →
17
44.8ms14% SOL
7.80×
Reported · license unknown · no source
2026-08-10

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
5.74ms
1.00×
Reported
license unknown · no source
unknown
6.10ms
1.06×
Reported
license unknown · no source
unknown
6.11ms
1.06×
Reported
license unknown · no source
unknown
6.32ms
1.10×
Reported
license unknown · no source
unknown
6.39ms
1.11×
Reported
license unknown · no source
unknown
10.6ms
1.85×
Reported
license unknown · no source
unknown
11.0ms
1.92×
Reported
license unknown · no source
unknown
11.9ms
2.08×
Reported
license unknown · no source
Show all 30 implementations ›
unknown
6.18ms
1.08×
Reported
license unknown · no source
unknown
6.19ms
1.08×
Reported
license unknown · no source
unknown
6.20ms
1.08×
Reported
license unknown · no source
unknown
6.40ms
1.12×
Reported
license unknown · no source
unknown
6.61ms
1.15×
Reported
license unknown · no source
unknown
6.63ms
1.15×
Reported
license unknown · no source
unknown
6.86ms
1.19×
Reported
license unknown · no source
unknown
7.15ms
1.25×
Reported
license unknown · no source
unknown
7.26ms
1.27×
Reported
license unknown · no source
unknown
7.58ms
1.32×
Reported
license unknown · no source
unknown
10.7ms
1.86×
Reported
license unknown · no source
unknown
10.8ms
1.88×
Reported
license unknown · no source
unknown
11.1ms
1.93×
Reported
license unknown · no source
unknown
11.5ms
2.01×
Reported
license unknown · no source
unknown
11.8ms
2.06×
Reported
license unknown · no source
unknown
12.1ms
2.11×
Reported
license unknown · no source
unknown
15.7ms
2.74×
Reported
license unknown · no source
unknown
16.3ms
2.83×
Reported
license unknown · no source
unknown
44.8ms
7.80×
Reported
license unknown · no source
unknown
5.83ms
1.02×
Reported
license unknown · no source
unknown
5.93ms
1.03×
Reported
license unknown · no source
unknown
6.09ms
1.06×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesbf16 [batch_size, seq_len, hidden_size]
gate_weightfp32 [num_experts, hidden_size]
e_score_correction_biasfp32 [1, num_experts]
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]
shared_gate_projbf16 [shared_intermediate_size, hidden_size]
shared_up_projbf16 [shared_intermediate_size, hidden_size]
shared_down_projbf16 [hidden_size, shared_intermediate_size]
norm_minfp32 scalar
outputbf16 [batch_size, seq_len, hidden_size]
router_logitsfp32 [batch_seq_len, num_experts]
Axes and behavior
top_kconstant = 8
seq_lenvariable
batch_sizevariable
hidden_sizeconstant = 8192
num_expertsconstant = 64
batch_seq_lenbatch_seq_len = batch_size * seq_len
intermediate_sizeconstant = 3584
shared_intermediate_sizeconstant = 7168
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias029_moe_sparse_routing_and_dispatchmodelernie-4-5-300b-a47b-ptsha256d5d9b1418a91…
Sources: NVIDIA SOL-ExecBench (2026-08-16)last observed 2026-08-16How records are decidedJSON