Skip to content
KernelIndex
Search⌘K

Chunk gated delta rule linear attention

29 eligible runs
other

Chunk-based Gated Delta Rule linear attention computation that processes sequences in chunks of 64 tokens. Implements delta rule update with exponential decay gating, chunk-wise attention with intra-chunk causal masking, inter-chunk recurrent state propagation, and cumulative decay tracking.

Fastest reported
45.2µsmean · 93% of SOL
RIAC as wellLicense 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 latencychunk_size = 64 · head_k_dim = 128 · head_v_dim = 128 · num_k_heads = 4 · num_v_heads = 1616 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 · 19 results · last observed 2026-08-22Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
45.2µs93% 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
54.6µs92% SOL
1.21×
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 →
3
57.7µs91% SOL
1.27×
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 →
4
59.3µs91% SOL
1.31×
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
60.1µs91% SOL
1.33×
Reported · license unknown · no source
2026-07-21

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

no sourcelicense unknownno install recipeRun detail →
6
60.4µs91% SOL
1.34×
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 →
7
63.5µs90% SOL
1.40×
Reported · license unknown · no source
2026-08-21

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

no sourcelicense unknownno install recipeRun detail →
8
64.9µs90% SOL
1.43×
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 →
9
65.2µs90% SOL
1.44×
Reported · license unknown · no source
2026-07-12

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

no sourcelicense unknownno install recipeRun detail →
10
72.5µs89% SOL
1.60×
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 →
11
72.9µs89% SOL
1.61×
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
89.2µs86% SOL
1.97×
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 →
13
94.2µs86% SOL
2.08×
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 →
14
97.1µs85% SOL
2.15×
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 →
15
126.9µs81% SOL
2.80×
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 →
16
274.2µs66% SOL
6.06×
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 →
17
597.2µs47% SOL
13.20×
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 →
18
1.49ms26% SOL
32.96×
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 →
19
5.22ms10% SOL
115.46×
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
59.3µs
1.31×
Reported
license unknown · no source
unknown
60.1µs
1.33×
Reported
license unknown · no source
unknown
65.2µs
1.44×
Reported
license unknown · no source
unknown
45.2µs
1.00×
Reported
license unknown · no source
unknown
54.6µs
1.21×
Reported
license unknown · no source
unknown
57.7µs
1.27×
Reported
license unknown · no source
unknown
63.5µs
1.40×
Reported
license unknown · no source
unknown
64.9µs
1.43×
Reported
license unknown · no source
Show all 29 implementations ›
unknown
70.4µs
1.56×
Reported
license unknown · no source
unknown
76.3µs
1.69×
Reported
license unknown · no source
unknown
525.4µs
11.61×
Reported
license unknown · no source
unknown
1.49ms
32.96×
Reported
license unknown · no source
unknown
4.72ms
104.36×
Reported
license unknown · no source
unknown
60.4µs
1.34×
Reported
license unknown · no source
unknown
72.5µs
1.60×
Reported
license unknown · no source
unknown
72.9µs
1.61×
Reported
license unknown · no source
unknown
89.2µs
1.97×
Reported
license unknown · no source
unknown
94.2µs
2.08×
Reported
license unknown · no source
unknown
97.1µs
2.15×
Reported
license unknown · no source
unknown
118.6µs
2.62×
Reported
license unknown · no source
unknown
124.5µs
2.75×
Reported
license unknown · no source
unknown
126.9µs
2.80×
Reported
license unknown · no source
unknown
274.2µs
6.06×
Reported
license unknown · no source
unknown
597.2µs
13.20×
Reported
license unknown · no source
unknown
1.52ms
33.60×
Reported
license unknown · no source
unknown
1.54ms
34.00×
Reported
license unknown · no source
unknown
2.05ms
45.32×
Reported
license unknown · no source
unknown
5.22ms
115.46×
Reported
license unknown · no source
unknown
5.78ms
127.75×
Reported
license unknown · no source

Semantics

Inputs and outputs
querybf16 [batch_size, num_k_heads, seq_len, head_k_dim]
keybf16 [batch_size, num_k_heads, seq_len, head_k_dim]
valuebf16 [batch_size, num_v_heads, seq_len, head_v_dim]
gbf16 [batch_size, num_v_heads, seq_len]
betabf16 [batch_size, num_v_heads, seq_len]
scalefp32 scalar
outputbf16 [batch_size, seq_len, num_v_heads, head_v_dim]
Axes and behavior
seq_lenvariable
batch_sizevariable
chunk_sizeconstant = 64
head_k_dimconstant = 128
head_v_dimconstant = 128
num_k_headsconstant = 4
num_v_headsconstant = 16
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias060_chunk_gated_delta_rule_linear_attentionmodelqwen3-next-80b-a3b-thinking-nvfp4sha2562ed516554b51…
Sources: NVIDIA SOL-ExecBench (2026-08-22)last observed 2026-08-22How records are decidedJSON