Skip to content
KernelIndex
Search⌘K

Vision cu seqlens variable length attention

34 eligible runs
attention

Vision encoder attention with variable-length sequence support using cumulative sequence lengths (cu_seqlens). Processes multiple images/videos with different resolutions in a single batch by using cu_seqlens to define sequence boundaries. Includes QKV projection, rotary position embeddings, scaled dot-product attention per sequence, and output projection.

Fastest reported
55.6µsmean · 91% of SOL
HyraLicense unknown · unknown

Reported evidence · last observed 2026-07-15. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhead_dim = 80 · num_heads = 16 · hidden_size = 1280 · qkv_out_size = 384016 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 · 21 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
55.6µs91% SOL
1.00×
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 →
2
58.6µs90% SOL
1.05×
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 →
3
64.4µs89% SOL
1.16×
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 →
4
66.9µs89% SOL
1.20×
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 →
5
74.5µs87% SOL
1.34×
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 →
6
81.2µs86% SOL
1.46×
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 →
7
85.9µs86% SOL
1.54×
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 →
8
88.6µs85% SOL
1.59×
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 →
9
88.9µs85% SOL
1.60×
Reported · license unknown · no source
2026-08-06

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

no sourcelicense unknownno install recipeRun detail →
10
89.5µs85% SOL
1.61×
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 →
11
90.2µs85% SOL
1.62×
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
93.3µs84% SOL
1.68×
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 →
13
94.7µs84% SOL
1.70×
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
100.5µs83% SOL
1.81×
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 →
15
115.7µs81% SOL
2.08×
Reported · license unknown · no source
2026-07-20

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

no sourcelicense unknownno install recipeRun detail →
16
167.1µs74% SOL
3.00×
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
253.0µs65% SOL
4.55×
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 →
18
360.7µs57% SOL
6.49×
Reported · license unknown · no source
2026-07-25

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

no sourcelicense unknownno install recipeRun detail →
19
395.5µs54% SOL
7.11×
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 →
20
399.4µs54% SOL
7.18×
Reported · license unknown · no source
2026-07-25

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

no sourcelicense unknownno install recipeRun detail →
21
508.0µs48% SOL
9.14×
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 →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
unknown
58.6µs
1.05×
Reported
license unknown · no source
unknown
71.1µs
1.28×
Reported
license unknown · no source
unknown
81.2µs
1.46×
Reported
license unknown · no source
unknown
86.7µs
1.56×
Reported
license unknown · no source
unknown
91.6µs
1.65×
Reported
license unknown · no source
unknown
115.7µs
2.08×
Reported
license unknown · no source
unknown
148.8µs
2.68×
Reported
license unknown · no source
unknown
360.9µs
6.49×
Reported
license unknown · no source
Show all 34 implementations ›
unknown
460.6µs
8.28×
Reported
license unknown · no source
unknown
718.9µs
12.93×
Reported
license unknown · no source
unknown
66.9µs
1.20×
Reported
license unknown · no source
unknown
74.5µs
1.34×
Reported
license unknown · no source
unknown
85.9µs
1.54×
Reported
license unknown · no source
unknown
88.6µs
1.59×
Reported
license unknown · no source
unknown
88.9µs
1.60×
Reported
license unknown · no source
unknown
89.5µs
1.61×
Reported
license unknown · no source
unknown
90.2µs
1.62×
Reported
license unknown · no source
unknown
90.5µs
1.63×
Reported
license unknown · no source
unknown
93.3µs
1.68×
Reported
license unknown · no source
unknown
94.7µs
1.70×
Reported
license unknown · no source
unknown
97.5µs
1.75×
Reported
license unknown · no source
unknown
100.5µs
1.81×
Reported
license unknown · no source
unknown
167.1µs
3.00×
Reported
license unknown · no source
unknown
253.0µs
4.55×
Reported
license unknown · no source
unknown
349.5µs
6.28×
Reported
license unknown · no source
unknown
360.7µs
6.49×
Reported
license unknown · no source
unknown
369.0µs
6.64×
Reported
license unknown · no source
unknown
395.5µs
7.11×
Reported
license unknown · no source
unknown
399.4µs
7.18×
Reported
license unknown · no source
unknown
508.0µs
9.14×
Reported
license unknown · no source
unknown
513.0µs
9.23×
Reported
license unknown · no source
unknown
55.6µs
1.00×
Reported
license unknown · no source
unknown
64.4µs
1.16×
Reported
license unknown · no source
unknown
69.6µs
1.25×
Reported
license unknown · no source

Semantics

Inputs and outputs
hidden_statesfp32 [total_seq_len, hidden_size]
cu_seqlensint64 [num_seqs]
cosfp32 [total_seq_len, head_dim]
sinfp32 [total_seq_len, head_dim]
qkv_weightfp32 [qkv_out_size, hidden_size]
qkv_biasfp32 [qkv_out_size]
proj_weightfp32 [hidden_size, hidden_size]
proj_biasfp32 [hidden_size]
outputfp32 [total_seq_len, hidden_size]
Axes and behavior
head_dimconstant = 80
num_seqsvariable
num_headsconstant = 16
hidden_sizeconstant = 1280
qkv_out_sizeqkv_out_size = hidden_size * 3
total_seq_lenvariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias021_vision_cu_seqlens_variable_length_attentionmodelqwen3-vl-8b-instructsha256a6663117340f…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON