Skip to content
KernelIndex
Search⌘K
Sign in

NVFP4 MoE expert linear with gating

40 eligible runs
moe

FP4-quantized MoE expert computation with SwiGLU gating. Performs output = down_proj(silu(gate_proj(x)) * up_proj(x)) where all projections use FP4 quantization with 16-element block-wise scaling.

Fastest reported
73.5µsmean · 95% of SOL
Hollow JellyfishLicense unknown · unknown

Reported evidence · last observed 2026-08-15. 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 · hidden_size_half = 1024 · intermediate_size = 5120 · hidden_scale_blocks = 128 · intermediate_size_half = 2560 · intermediate_scale_blocks = 32016 cases
num_tokens
dtype
128bf16/fp4/fp8_e4m3256bf16/fp4/fp8_e4m3384bf16/fp4/fp8_e4m3896bf16/fp4/fp8_e4m31024bf16/fp4/fp8_e4m31408bf16/fp4/fp8_e4m31792bf16/fp4/fp8_e4m31920bf16/fp4/fp8_e4m32176bf16/fp4/fp8_e4m32560bf16/fp4/fp8_e4m32688bf16/fp4/fp8_e4m32944bf16/fp4/fp8_e4m33328bf16/fp4/fp8_e4m33456bf16/fp4/fp8_e4m33840bf16/fp4/fp8_e4m33968bf16/fp4/fp8_e4m3
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 173.5 µsHollow Jellyfish27NVIDIA B200 · env 2121.0 µsRecursive13

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 · 27 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
Hollow Jellyfish
73.5µs95% SOL
1.00×
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 →
2
Amir M. Mir | SF Tensor
90.3µs93% SOL
1.23×
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 →
3
Geometric
98.7µs93% SOL
1.34×
Reported · license unknown · no source
2026-08-24

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

no sourcelicense unknownno install recipeRun detail →
4
Amir M. Mir | SF Tensor
99.5µs92% SOL
1.35×
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 →
5
Geometric
99.9µs92% SOL
1.36×
Reported · license unknown · no source
2026-08-22

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

no sourcelicense unknownno install recipeRun detail →
6
Yatimai
102.7µs92% SOL
1.40×
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 →
7
Geometric
108.5µs92% SOL
1.48×
Reported · license unknown · no source
2026-08-18

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

no sourcelicense unknownno install recipeRun detail →
8
Geometric
109.0µs92% SOL
1.48×
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 →
9
Geometric
114.9µs91% SOL
1.56×
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 →
10
Geometric
115.4µs91% SOL
1.57×
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 →
11
ac4k
117.4µs91% SOL
1.60×
Reported · license unknown · no source
2026-08-02

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

no sourcelicense unknownno install recipeRun detail →
12
Yatimai
117.5µs91% SOL
1.60×
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 →
13
Gleaming Coyote
118.0µs91% SOL
1.60×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
14
Electric Cheetah
118.0µs91% SOL
1.61×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
15
Rapid Tiger
118.5µs91% SOL
1.61×
Reported · license unknown · no source
2026-08-14

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

no sourcelicense unknownno install recipeRun detail →
16
Recursive
119.5µs91% SOL
1.63×
Reported · license unknown · no source
2026-07-14

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

no sourcelicense unknownno install recipeRun detail →
17
Geometric
126.1µs90% SOL
1.72×
Reported · license unknown · no source
2026-08-03

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

no sourcelicense unknownno install recipeRun detail →
18
Geometric
134.1µs90% SOL
1.83×
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 →
19
Yatimai
135.7µs90% SOL
1.85×
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 →
20
GlugGlug
137.3µs89% SOL
1.87×
Reported · license unknown · no source
2026-07-14

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

no sourcelicense unknownno install recipeRun detail →
21
Chrome Lion
141.9µs89% SOL
1.93×
Reported · license unknown · no source
2026-07-14

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

no sourcelicense unknownno install recipeRun detail →
22
doubleAI
162.4µs87% SOL
2.21×
Reported · license unknown · no source
2026-07-13

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

no sourcelicense unknownno install recipeRun detail →
23
Hyra
166.1µs87% SOL
2.26×
Reported · license unknown · no source
2026-07-15

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

no sourcelicense unknownno install recipeRun detail →
24
Geometric
177.4µs87% SOL
2.41×
Reported · license unknown · no source
2026-08-03

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

no sourcelicense unknownno install recipeRun detail →
25
Amir M. Mir | SF Tensor
206.6µs85% SOL
2.81×
Reported · license unknown · no source
2026-07-13

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

no sourcelicense unknownno install recipeRun detail →
26
Geometric
391.3µs74% SOL
5.32×
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 →
27
Geometric
623.8µs63% SOL
8.49×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
Hollow Jellyfish
unknown
73.5µs
1.00×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
90.3µs
1.23×
Reported
license unknown · no source
Yatimai
unknown
102.7µs
1.40×
Reported
license unknown · no source
Geometric
unknown
98.7µs
1.34×
Reported
license unknown · no source
ac4k
unknown
117.4µs
1.60×
Reported
license unknown · no source
Gleaming Coyote
unknown
118.0µs
1.60×
Reported
license unknown · no source
Electric Cheetah
unknown
118.0µs
1.61×
Reported
license unknown · no source
Rapid Tiger
unknown
118.5µs
1.61×
Reported
license unknown · no source
Show all 40 implementations ›
Recursive
unknown
119.5µs
1.63×
Reported
license unknown · no source
GlugGlug
unknown
136.1µs
1.85×
Reported
license unknown · no source
Chrome Lion
unknown
139.4µs
1.90×
Reported
license unknown · no source
doubleAI
unknown
154.8µs
2.11×
Reported
license unknown · no source
Hyra
unknown
166.1µs
2.26×
Reported
license unknown · no source
Cursor
unknown
905.2µs
12.32×
Reported
license unknown · no source
Geometric
unknown
99.9µs
1.36×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
99.5µs
1.35×
Reported
license unknown · no source
Geometric
unknown
108.5µs
1.48×
Reported
license unknown · no source
Geometric
unknown
109.0µs
1.48×
Reported
license unknown · no source
Yatimai
unknown
117.5µs
1.60×
Reported
license unknown · no source
Geometric
unknown
115.4µs
1.57×
Reported
license unknown · no source
Geometric
unknown
114.9µs
1.56×
Reported
license unknown · no source
Recursive
unknown
121.0µs
1.65×
Reported
license unknown · no source
Geometric
unknown
126.1µs
1.72×
Reported
license unknown · no source
Geometric
unknown
134.1µs
1.83×
Reported
license unknown · no source
Yatimai
unknown
135.7µs
1.85×
Reported
license unknown · no source
GlugGlug
unknown
136.4µs
1.86×
Reported
license unknown · no source
GlugGlug
unknown
136.6µs
1.86×
Reported
license unknown · no source
GlugGlug
unknown
136.9µs
1.86×
Reported
license unknown · no source
GlugGlug
unknown
137.0µs
1.86×
Reported
license unknown · no source
GlugGlug
unknown
137.3µs
1.87×
Reported
license unknown · no source
GlugGlug
unknown
139.4µs
1.90×
Reported
license unknown · no source
Chrome Lion
unknown
141.9µs
1.93×
Reported
license unknown · no source
doubleAI
unknown
162.4µs
2.21×
Reported
license unknown · no source
Geometric
unknown
177.4µs
2.41×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
204.2µs
2.78×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
206.6µs
2.81×
Reported
license unknown · no source
Geometric
unknown
391.3µs
5.32×
Reported
license unknown · no source
doubleAI
unknown
419.3µs
5.70×
Reported
license unknown · no source
Geometric
unknown
623.8µs
8.49×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
917.8µs
12.49×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesbf16 [num_tokens, hidden_size]
gate_weight_fp4float4_e2m1fn_x2 [1, intermediate_size, hidden_size_half]
gate_scalefp8_e4m3 [1, intermediate_size, hidden_scale_blocks]
gate_global_decodefp32 scalar
up_weight_fp4float4_e2m1fn_x2 [1, intermediate_size, hidden_size_half]
up_scalefp8_e4m3 [1, intermediate_size, hidden_scale_blocks]
up_global_decodefp32 scalar
down_weight_fp4float4_e2m1fn_x2 [1, hidden_size, intermediate_size_half]
down_scalefp8_e4m3 [1, hidden_size, intermediate_scale_blocks]
down_global_decodefp32 scalar
outputbf16 [num_tokens, hidden_size]
Axes and behavior
num_tokensvariable
hidden_sizeconstant = 2048
hidden_size_halfhidden_size_half = hidden_size // 2
intermediate_sizeconstant = 5120
hidden_scale_blockshidden_scale_blocks = hidden_size // 16
intermediate_size_halfintermediate_size_half = intermediate_size // 2
intermediate_scale_blocksintermediate_scale_blocks = intermediate_size // 16
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias032_nvfp4_moe_expert_linear_with_gatingmodelqwen3-next-80b-a3b-thinking-nvfp4sha2569cca569da97f…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON