Seqlen-finetuned-reconstructed Hyena complete forward block

41 eligible runs

ssm

Complete end-to-end HyenaBlock forward pass that fuses all operations from input to output: residual management, layer normalizations with dtype conversions, the full Hyena operator (input projection, short conv, filter generation, FFT convolution, gating), MLP with GELU activation, and final residual addition.

Fastest reported

21.5µsmean · 97% of SOL

HyraLicense unknown · unknown

Reported evidence · last observed 2026-07-17. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyl_max = 32768 · order = 2 · d_inner = 1024 · d_model = 256 · emb_dim = 5 · inner_width = 768 · filter_order = 64 · order_minus_1 = 1 · short_filter_order = 316 cases

seq_len

batch_size

dtype

1282fp32 1311fp32 2561fp32 2564fp32 25632fp32 5121fp32 5122fp32 51216fp32 5414fp32 9978fp32 10241fp32 20481fp32 20482fp32 20532fp32 40961fp32 40962fp32

scrolls · 16 cases total

Aggregatessuite of 16 cases · mean latency

HardwareBest knownImplementationRuns

NVIDIA B200 · env 121.5 µsHyra32 NVIDIA B200 · env 2105.7 µsdoubleAI9

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 · 32 results · last observed 2026-08-14Record history →

#

Implementation

Latency

vs #1

Trust

Observed

1

21.5µs97% SOL

1.00×

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 →

2

Amir M. Mir | SF Tensor

26.0µs96% SOL

1.21×

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 →

3

26.1µs96% SOL

1.22×

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 →

4

26.3µs96% SOL

1.22×

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 →

5

Gleaming Coyote

26.3µs96% SOL

1.23×

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

26.8µs96% SOL

1.25×

Reported · license unknown · no source

2026-08-04

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

no sourcelicense unknownno install recipeRun detail →

7

Amir M. Mir | SF Tensor

45.4µs92% SOL

2.11×

Reported · license unknown · no source

2026-08-04

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

no sourcelicense unknownno install recipeRun detail →

8

49.1µs92% SOL

2.29×

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 →

9

Proud Wren @ NVIDIA

101.2µs84% SOL

4.71×

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 →

10

109.8µs83% SOL

5.11×

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 →

11

Amir M. Mir | SF Tensor

114.4µs82% SOL

5.33×

Reported · license unknown · no source

2026-07-26

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

no sourcelicense unknownno install recipeRun detail →

12

133.1µs80% SOL

6.19×

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 →

13

140.7µs79% SOL

6.55×

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

141.8µs79% SOL

6.60×

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 →

15

Amir M. Mir | SF Tensor

149.6µs78% SOL

6.96×

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 →

16

168.2µs76% SOL

7.83×

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 →

17

170.0µs76% SOL

7.91×

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 →

18

186.5µs74% SOL

8.68×

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 →

19

Amir M. Mir | SF Tensor

201.0µs72% SOL

9.36×

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 →

20

272.0µs66% SOL

12.66×

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 →

21

296.8µs64% SOL

13.82×

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 →

22

397.8µs57% SOL

18.52×

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 →

23

410.2µs56% SOL

19.10×

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 →

24

424.2µs55% SOL

19.75×

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 →

25

425.2µs55% SOL

19.80×

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 →

26

471.0µs52% SOL

21.93×

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 →

27

514.4µs50% SOL

23.95×

Reported · license unknown · no source

2026-07-26

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

no sourcelicense unknownno install recipeRun detail →

28

543.1µs49% SOL

25.29×

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 →

29

614.1µs46% SOL

28.59×

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 →

30

790.0µs40% SOL

36.78×

Reported · license unknown · no source

2026-07-26

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

no sourcelicense unknownno install recipeRun detail →

31

895.8µs37% SOL

41.71×

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 →

32

963.7µs35% SOL

44.87×

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 →

Implementations

Implementation

Runtime

Best latency

Evidence

Availability

unknown

21.5µs

1.00×

Reported

license unknown · no source

Gleaming Coyote

unknown

26.3µs

1.23×

Reported

license unknown · no source

unknown

26.8µs

1.25×

Reported

license unknown · no source

unknown

49.1µs

2.29×

Reported

license unknown · no source

Proud Wren @ NVIDIA

unknown

101.2µs

4.71×

Reported

license unknown · no source

unknown

105.7µs

4.92×

Reported

license unknown · no source

unknown

135.2µs

6.29×

Reported

license unknown · no source

unknown

168.2µs

7.83×

Reported

license unknown · no source

Show all 41 implementations ›

unknown

387.9µs

18.06×

Reported

license unknown · no source

unknown

550.0µs

25.61×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

45.4µs

2.11×

Reported

license unknown · no source

unknown

109.8µs

5.11×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

114.4µs

5.33×

Reported

license unknown · no source

unknown

133.1µs

6.19×

Reported

license unknown · no source

unknown

140.7µs

6.55×

Reported

license unknown · no source

unknown

141.8µs

6.60×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

149.6µs

6.96×

Reported

license unknown · no source

unknown

170.0µs

7.91×

Reported

license unknown · no source

unknown

186.5µs

8.68×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

199.1µs

9.27×

Reported

license unknown · no source

unknown

200.6µs

9.34×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

201.0µs

9.36×

Reported

license unknown · no source

unknown

272.0µs

12.66×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

286.3µs

13.33×

Reported

license unknown · no source

unknown

296.8µs

13.82×

Reported

license unknown · no source

unknown

397.8µs

18.52×

Reported

license unknown · no source

unknown

410.2µs

19.10×

Reported

license unknown · no source

unknown

424.2µs

19.75×

Reported

license unknown · no source

unknown

425.2µs

19.80×

Reported

license unknown · no source

unknown

471.0µs

21.93×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

481.6µs

22.42×

Reported

license unknown · no source

unknown

514.4µs

23.95×

Reported

license unknown · no source

unknown

543.1µs

25.29×

Reported

license unknown · no source

unknown

550.9µs

25.65×

Reported

license unknown · no source

unknown

614.1µs

28.59×

Reported

license unknown · no source

unknown

790.0µs

36.78×

Reported

license unknown · no source

unknown

895.8µs

41.71×

Reported

license unknown · no source

unknown

963.7µs

44.87×

Reported

license unknown · no source

Amir M. Mir | SF Tensor

unknown

26.0µs

1.21×

Reported

license unknown · no source

unknown

26.1µs

1.22×

Reported

license unknown · no source

unknown

26.3µs

1.22×

Reported

license unknown · no source

Semantics

Inputs and outputs

hidden_statesfp32 [batch_size, seq_len, d_model]

norm1_weightfp32 [d_model]

norm1_biasfp32 [d_model]

norm2_weightfp32 [d_model]

norm2_biasfp32 [d_model]

in_proj_weightfp32 [inner_width, d_model]

in_proj_biasfp32 [inner_width]

short_conv_weightfp32 [inner_width, 1, short_filter_order]

short_conv_biasfp32 [inner_width]

filter_linear1_weightfp32 [filter_order, emb_dim]

filter_linear1_biasfp32 [filter_order]

sin_freqfp32 [1, filter_order]

filter_linear2_weightfp32 [filter_order, filter_order]

filter_linear2_biasfp32 [filter_order]

filter_linear3_weightfp32 [filter_order, filter_order]

filter_linear3_biasfp32 [filter_order]

filter_linear_final_weightfp32 [d_model, filter_order]

filter_biasfp32 [d_model]

exp_mod_deltasfp32 [1, 1, d_model]

out_proj_weightfp32 [d_model, d_model]

out_proj_biasfp32 [d_model]

mlp_fc1_weightfp32 [d_inner, d_model]

mlp_fc1_biasfp32 [d_inner]

mlp_fc2_weightfp32 [d_model, d_inner]

mlp_fc2_biasfp32 [d_model]

layer_norm_epsfp32 scalar

exp_mod_shiftfp32 scalar

outputfp32 [batch_size, seq_len, d_model]

Axes and behavior

l_maxconstant = 32768

orderconstant = 2

d_innerconstant = 1024

d_modelconstant = 256

emb_dimconstant = 5

seq_lenvariable

batch_sizevariable

inner_widthinner_width = d_model * (order + 1)

filter_orderconstant = 64

order_minus_1order_minus_1 = order - 1

short_filter_orderconstant = 3

determinismunspecified

constraintsNo mutation or aliasing

Identity

alias051_seqlen-finetuned-reconstructed_hyena_complete_forward_blockmodelhyena-small-32ksha256e1a03aa3a33e…

Sources: NVIDIA SOL-ExecBench (2026-08-14)last observed 2026-08-14How records are decided JSON