Skip to content
KernelIndex
Search⌘K
Sign in

NVFP4 fused MLP SiLU gating

29 eligible runs
moe

FP4-quantized fused MLP with SiLU activation and gating mechanism. Implements down_proj(SiLU(gate_proj(x)) * up_proj(x)) with FP4 quantization and block-wise scaling (16 elements per block). Gate and up projections share input quantization, followed by fused SiLU gating, then down projection.

Fastest reported
208.2µsmean · 91% of SOL
Hollow JellyfishLicense 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_half = 2048 · hidden_size = 4096 · intermediate_half = 6144 · intermediate_size = 12288 · hidden_scale_blocks = 256 · intermediate_scale_blocks = 76816 cases
seq_len
batch_size
dtype
1281bf16/fp4/fp8_e4m31282bf16/fp4/fp8_e4m31284bf16/fp4/fp8_e4m31288bf16/fp4/fp8_e4m312832bf16/fp4/fp8_e4m32561bf16/fp4/fp8_e4m32564bf16/fp4/fp8_e4m35121bf16/fp4/fp8_e4m35122bf16/fp4/fp8_e4m35128bf16/fp4/fp8_e4m351232bf16/fp4/fp8_e4m310242bf16/fp4/fp8_e4m320481bf16/fp4/fp8_e4m320484bf16/fp4/fp8_e4m320488bf16/fp4/fp8_e4m340961bf16/fp4/fp8_e4m3
scrolls · 16 cases total
Aggregatessuite of 16 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 1208.2 µsHollow Jellyfish23NVIDIA B200 · env 2230.0 µsRecursive6

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 · 23 results · last observed 2026-08-20Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
Hollow Jellyfish
208.2µs91% 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
Amir M. Mir | SF Tensor
210.0µs91% SOL
1.01×
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 →
3
ac4k
218.4µs90% SOL
1.05×
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 →
4
Amir M. Mir | SF Tensor
221.6µs90% SOL
1.06×
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 →
5
Electric Cheetah
229.6µs90% SOL
1.10×
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 →
6
Rapid Tiger
231.6µs89% SOL
1.11×
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 →
7
Rapid Tiger
232.0µs89% SOL
1.11×
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 →
8
Gleaming Coyote
236.5µs89% SOL
1.14×
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 →
9
Recursive
236.9µs89% SOL
1.14×
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 →
10
Hyra
255.6µs88% SOL
1.23×
Reported · license unknown · no source
2026-07-18

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

no sourcelicense unknownno install recipeRun detail →
11
Geometric
259.3µs88% SOL
1.25×
Reported · license unknown · no source
2026-08-17

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

no sourcelicense unknownno install recipeRun detail →
12
Geometric
267.2µs88% SOL
1.28×
Reported · license unknown · no source
2026-08-17

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

no sourcelicense unknownno install recipeRun detail →
13
Q H
273.1µs88% SOL
1.31×
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 →
14
Geometric
295.1µs87% SOL
1.42×
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 →
15
Geometric
302.2µs86% SOL
1.45×
Reported · license unknown · no source
2026-08-12

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

no sourcelicense unknownno install recipeRun detail →
16
Geometric
318.9µs86% SOL
1.53×
Reported · license unknown · no source
2026-08-09

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

no sourcelicense unknownno install recipeRun detail →
17
Yatimai
340.6µs85% SOL
1.64×
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 →
18
doubleAI
393.8µs83% SOL
1.89×
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 →
19
Geometric
473.6µs80% SOL
2.27×
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 →
20
Geometric
603.6µs74% SOL
2.90×
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 →
21
Chrome Lion
816.8µs68% SOL
3.92×
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
Cursor
1.55ms53% SOL
7.44×
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
Geometric
5.34ms24% SOL
25.67×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
Amir M. Mir | SF Tensor
unknown
210.0µs
1.01×
Reported
license unknown · no source
ac4k
unknown
218.4µs
1.05×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
221.6µs
1.06×
Reported
license unknown · no source
Hollow Jellyfish
unknown
208.2µs
1.00×
Reported
license unknown · no source
Recursive
unknown
230.0µs
1.10×
Reported
license unknown · no source
Electric Cheetah
unknown
229.6µs
1.10×
Reported
license unknown · no source
Rapid Tiger
unknown
231.6µs
1.11×
Reported
license unknown · no source
Gleaming Coyote
unknown
236.5µs
1.14×
Reported
license unknown · no source
Show all 29 implementations ›
Hyra
unknown
255.6µs
1.23×
Reported
license unknown · no source
Geometric
unknown
259.3µs
1.25×
Reported
license unknown · no source
Q H
unknown
273.1µs
1.31×
Reported
license unknown · no source
doubleAI
unknown
291.8µs
1.40×
Reported
license unknown · no source
Yatimai
unknown
340.6µs
1.64×
Reported
license unknown · no source
Chrome Lion
unknown
807.1µs
3.88×
Reported
license unknown · no source
Cursor
unknown
1.55ms
7.44×
Reported
license unknown · no source
Rapid Tiger
unknown
232.0µs
1.11×
Reported
license unknown · no source
Recursive
unknown
236.9µs
1.14×
Reported
license unknown · no source
Geometric
unknown
267.2µs
1.28×
Reported
license unknown · no source
Geometric
unknown
295.1µs
1.42×
Reported
license unknown · no source
Geometric
unknown
302.2µs
1.45×
Reported
license unknown · no source
Geometric
unknown
318.9µs
1.53×
Reported
license unknown · no source
doubleAI
unknown
393.8µs
1.89×
Reported
license unknown · no source
doubleAI
unknown
460.2µs
2.21×
Reported
license unknown · no source
Geometric
unknown
473.6µs
2.27×
Reported
license unknown · no source
Geometric
unknown
603.6µs
2.90×
Reported
license unknown · no source
Chrome Lion
unknown
816.8µs
3.92×
Reported
license unknown · no source
Cursor
unknown
1.58ms
7.58×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
2.00ms
9.61×
Reported
license unknown · no source
Geometric
unknown
5.34ms
25.67×
Reported
license unknown · no source

Semantics

Inputs and outputs
xbf16 [batch_size, seq_len, hidden_size]
gate_weight_fp4float4_e2m1fn_x2 [1, intermediate_size, hidden_half]
gate_scalefp8_e4m3 [1, intermediate_size, hidden_scale_blocks]
gate_global_decode_scalefp32 scalar
up_weight_fp4float4_e2m1fn_x2 [1, intermediate_size, hidden_half]
up_scalefp8_e4m3 [1, intermediate_size, hidden_scale_blocks]
up_global_decode_scalefp32 scalar
down_weight_fp4float4_e2m1fn_x2 [1, hidden_size, intermediate_half]
down_scalefp8_e4m3 [1, hidden_size, intermediate_scale_blocks]
down_global_decode_scalefp32 scalar
outputbf16 [batch_size, seq_len, hidden_size]
Axes and behavior
seq_lenvariable
batch_sizevariable
hidden_halfhidden_half = hidden_size // 2
hidden_sizeconstant = 4096
intermediate_halfintermediate_half = intermediate_size // 2
intermediate_sizeconstant = 12288
hidden_scale_blockshidden_scale_blocks = hidden_size // 16
intermediate_scale_blocksintermediate_scale_blocks = intermediate_size // 16
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias029_nvfp4_fused_mlp_silu_gatingmodelqwen3-8b-nvfp4sha2560dacd7e3013e…
Sources: NVIDIA SOL-ExecBench (2026-08-20)last observed 2026-08-20How records are decidedJSON