Skip to content
KernelIndex
Search⌘K
Sign in

MLA paged decode h16 ckv512 kpe64 ps1

281 eligible runs
mla-paged-attention

Batched Multi-head Latent Attention decode with a paged KV cache. Captured from DeepSeek-V3 with tensor parallel size 8.

Fastest reported
13.2µsmean · 97% of SOL
RIAC as wellLicense unknown · unknown

Reported evidence · last observed 2026-08-10. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 47 cases · mean latencynum_pages = 98966994 cases
num_kv_indices
batch_size
len_indptr
head_dim_ckv
head_dim_kpe
num_qo_heads
page_size
dtype
812————bf1681251264161bf1610812————bf161081251264161bf1620812————bf162081251264161bf1630812————bf163081251264161bf1640812————bf164081251264161bf164571617————bf16457161751264161bf1650812————bf165081251264161bf1660812————bf166081251264161bf1670812————bf167081251264161bf1680812————bf168081251264161bf16100812————bf1610081251264161bf16110812————bf1611081251264161bf16120812————bf1612081251264161bf1618571617————bf161857161751264161bf16190812————bf1619081251264161bf16240812————bf1624081251264161bf16270812————bf1627081251264161bf1627571617————bf162757161751264161bf1635571617————bf163557161751264161bf1643571617————bf164357161751264161bf1650571617————bf165057161751264161bf1658571617————bf165857161751264161bf1666571617————bf166657161751264161bf1672571617————bf167257161751264161bf1680571617————bf168057161751264161bf1688571617————bf168857161751264161bf1696571617————bf169657161751264161bf1699456465————bf169945646551264161bf16108571617————bf1610857161751264161bf16128571617————bf1612857161751264161bf16148571617————bf1614857161751264161bf16163456465————bf1616345646551264161bf16172571617————bf1617257161751264161bf16227456465————bf1622745646551264161bf16275456465————bf1627545646551264161bf16307456465————bf1630745646551264161bf16339456465————bf1633945646551264161bf16371456465————bf1637145646551264161bf16403456465————bf1640345646551264161bf16448456465————bf1644845646551264161bf16480456465————bf1648045646551264161bf16512456465————bf1651245646551264161bf16544456465————bf1654445646551264161bf16576456465————bf1657645646551264161bf16623456465————bf1662345646551264161bf16687456465————bf1668745646551264161bf16751456465————bf1675145646551264161bf16
scrolls · 94 cases total
Aggregatessuite of 47 cases · mean latency
HardwareBest knownImplementationRuns
NVIDIA B200 · env 113.2 µsRIAC as well62NVIDIA B200 · env 222.4 µsPhantom Orca31

Not measured on H100 for this workload. Challenges →

Source-native comparison · GPU NVIDIA B200 · Workload suite of 47 cases · mean latency · Protocol SOL-ExecBench evaluation stack v1.1 · mean · 62 results · last observed 2026-08-18Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
RIAC as well
13.2µs97% SOL
1.00×
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 →
2
Amir M. Mir | SF Tensor
13.7µs97% SOL
1.04×
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 →
3
Amir M. Mir | SF Tensor
13.9µs97% SOL
1.06×
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 →
4
Amir M. Mir | SF Tensor
14.3µs97% SOL
1.09×
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 →
5
Amir M. Mir | SF Tensor
15.0µs96% SOL
1.14×
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 →
6
Geometric
15.2µs96% SOL
1.16×
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 →
7
Geometric
16.3µs96% SOL
1.24×
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
Amir M. Mir | SF Tensor
16.3µs96% SOL
1.24×
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 →
9
Amir M. Mir | SF Tensor
16.7µs96% SOL
1.27×
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
16.8µs96% SOL
1.28×
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 →
11
Geometric
16.9µs96% SOL
1.29×
Reported · license unknown · no source
2026-08-13

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

no sourcelicense unknownno install recipeRun detail →
12
Amir M. Mir | SF Tensor
17.1µs96% SOL
1.30×
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 →
13
Amir M. Mir | SF Tensor
17.1µs96% SOL
1.30×
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 →
14
Vibe Kernel
17.3µs96% SOL
1.31×
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 →
15
Amir M. Mir | SF Tensor
17.3µs96% SOL
1.31×
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 →
16
Amir M. Mir | SF Tensor
17.5µs96% SOL
1.33×
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 →
17
Geometric
17.8µs96% SOL
1.35×
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 →
18
Stellar Flamingo
18.0µs96% SOL
1.37×
Reported · license unknown · no source
2026-08-13

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

no sourcelicense unknownno install recipeRun detail →
19
Hyra
18.3µs96% SOL
1.39×
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 →
20
Amir M. Mir | SF Tensor
18.5µs95% SOL
1.40×
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 →
21
Amir M. Mir | SF Tensor
18.6µs95% SOL
1.41×
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 →
22
Q H
18.9µs95% SOL
1.44×
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 →
23
Vibe Kernel
19.5µs95% SOL
1.48×
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 →
24
Stellar Flamingo
19.6µs95% SOL
1.49×
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 →
25
Amir M. Mir | SF Tensor
19.9µs95% SOL
1.51×
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 →
26
Stellar Flamingo
20.5µs95% SOL
1.56×
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 →
27
Q H
21.5µs95% SOL
1.63×
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 →
28
Q H
21.5µs95% SOL
1.63×
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 →
29
Qwen
22.0µs95% SOL
1.67×
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 →
30
Phantom Orca
23.0µs94% SOL
1.75×
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 →
31
ac4k
23.2µs94% SOL
1.76×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
32
ac4k
23.3µs94% SOL
1.77×
Reported · license unknown · no source
2026-07-31

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

no sourcelicense unknownno install recipeRun detail →
33
Vibe Kernel
23.4µs94% SOL
1.78×
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 →
34
AKO
23.6µs94% SOL
1.79×
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 →
35
ac4k
23.6µs94% SOL
1.79×
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 →
36
Geometric
24.7µs94% SOL
1.88×
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 →
37
Geometric
24.8µs94% SOL
1.88×
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 →
38
Q H
24.9µs94% SOL
1.89×
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 →
39
Recursive
25.0µs94% SOL
1.90×
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 →
40
Geometric
25.1µs94% SOL
1.91×
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 →
41
Geometric
25.2µs94% SOL
1.91×
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 →
42
Stellar Flamingo
25.3µs94% SOL
1.92×
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 →
43
doubleAI
25.5µs94% SOL
1.94×
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 →
44
Geometric
25.9µs94% SOL
1.97×
Reported · license unknown · no source
2026-07-31

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

no sourcelicense unknownno install recipeRun detail →
45
Geometric
26.1µs94% SOL
1.98×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
46
Geometric
26.4µs94% SOL
2.01×
Reported · license unknown · no source
2026-08-01

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

no sourcelicense unknownno install recipeRun detail →
47
Coral@HumanAgentSociety
26.7µs94% SOL
2.03×
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 →
48
Q H
27.0µs93% SOL
2.05×
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 →
49
Geometric
27.9µs88% SOL
2.12×
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 →
50
Geometric
28.8µs93% SOL
2.19×
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 →
51
Stellar Flamingo
29.5µs93% SOL
2.24×
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 →
52
Geometric
30.2µs93% SOL
2.29×
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 →
53
Stellar Flamingo
32.7µs92% SOL
2.48×
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 →
54
Stellar Flamingo
32.8µs92% SOL
2.49×
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 →
55
Stellar Flamingo
33.8µs92% SOL
2.57×
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 →
56
Stellar Flamingo
34.0µs92% SOL
2.58×
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 →
57
Geometric
41.9µs90% SOL
3.18×
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 →
58
Geometric
47.9µs81% SOL
3.64×
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 →
59
Amir M. Mir | SF Tensor
49.6µs88% SOL
3.77×
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 →
60
Geometric
97.5µs80% SOL
7.40×
Reported · license unknown · no source
2026-07-19

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

no sourcelicense unknownno install recipeRun detail →

2 more rows in this comparison group. Open all in search →

Implementations

Implementation
Runtime
Best latency
Evidence
Availability
Amir M. Mir | SF Tensor
unknown
13.7µs
1.04×
Reported
license unknown · no source
Geometric
unknown
15.2µs
1.16×
Reported
license unknown · no source
Vibe Kernel
unknown
17.3µs
1.31×
Reported
license unknown · no source
Stellar Flamingo
unknown
18.0µs
1.37×
Reported
license unknown · no source
Hyra
unknown
18.3µs
1.39×
Reported
license unknown · no source
Q H
unknown
18.9µs
1.44×
Reported
license unknown · no source
Qwen
unknown
22.0µs
1.67×
Reported
license unknown · no source
doubleAI
unknown
22.9µs
1.74×
Reported
license unknown · no source
Show all 102 implementations ›
Phantom Orca
unknown
22.4µs
1.70×
Reported
license unknown · no source
ac4k
unknown
23.2µs
1.76×
Reported
license unknown · no source
AKO
unknown
23.6µs
1.79×
Reported
license unknown · no source
Recursive
unknown
24.3µs
1.85×
Reported
license unknown · no source
Coral@HumanAgentSociety
unknown
26.0µs
1.98×
Reported
license unknown · no source
AKO4ALL
unknown
31.2µs
2.37×
Reported
license unknown · no source
Argus
unknown
33.6µs
2.55×
Reported
license unknown · no source
IcySky
unknown
34.5µs
2.62×
Reported
license unknown · no source
Bold Tiger
unknown
41.4µs
3.15×
Reported
license unknown · no source
KerSor
unknown
77.4µs
5.87×
Reported
license unknown · no source
Cursor
unknown
160.4µs
12.18×
Reported
license unknown · no source
Nimble Penguin
unknown
195.3µs
14.83×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
13.9µs
1.06×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
14.3µs
1.09×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
15.0µs
1.14×
Reported
license unknown · no source
Geometric
unknown
16.3µs
1.24×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
16.3µs
1.24×
Reported
license unknown · no source
Geometric
unknown
16.8µs
1.28×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
16.7µs
1.27×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
17.1µs
1.30×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
17.1µs
1.30×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
17.3µs
1.31×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
17.5µs
1.33×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
18.5µs
1.40×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
18.6µs
1.41×
Reported
license unknown · no source
Stellar Flamingo
unknown
19.6µs
1.49×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
19.9µs
1.51×
Reported
license unknown · no source
Stellar Flamingo
unknown
20.5µs
1.56×
Reported
license unknown · no source
Vibe Kernel
unknown
19.5µs
1.48×
Reported
license unknown · no source
Q H
unknown
21.5µs
1.63×
Reported
license unknown · no source
Q H
unknown
21.5µs
1.63×
Reported
license unknown · no source
ac4k
unknown
23.3µs
1.77×
Reported
license unknown · no source
Phantom Orca
unknown
23.0µs
1.75×
Reported
license unknown · no source
ac4k
unknown
23.6µs
1.79×
Reported
license unknown · no source
Phantom Orca
unknown
23.1µs
1.76×
Reported
license unknown · no source
Geometric
unknown
24.8µs
1.88×
Reported
license unknown · no source
Vibe Kernel
unknown
23.4µs
1.78×
Reported
license unknown · no source
Geometric
unknown
24.7µs
1.88×
Reported
license unknown · no source
Geometric
unknown
25.1µs
1.91×
Reported
license unknown · no source
Geometric
unknown
25.2µs
1.91×
Reported
license unknown · no source
Recursive
unknown
25.0µs
1.90×
Reported
license unknown · no source
doubleAI
unknown
25.5µs
1.94×
Reported
license unknown · no source
Q H
unknown
24.9µs
1.89×
Reported
license unknown · no source
Geometric
unknown
25.9µs
1.97×
Reported
license unknown · no source
Stellar Flamingo
unknown
25.3µs
1.92×
Reported
license unknown · no source
Geometric
unknown
26.1µs
1.98×
Reported
license unknown · no source
Geometric
unknown
26.4µs
2.01×
Reported
license unknown · no source
Coral@HumanAgentSociety
unknown
26.7µs
2.03×
Reported
license unknown · no source
Q H
unknown
27.0µs
2.05×
Reported
license unknown · no source
Geometric
unknown
28.8µs
2.19×
Reported
license unknown · no source
doubleAI
unknown
29.5µs
2.24×
Reported
license unknown · no source
AKO
unknown
29.6µs
2.25×
Reported
license unknown · no source
Geometric
unknown
30.2µs
2.29×
Reported
license unknown · no source
Stellar Flamingo
unknown
29.5µs
2.24×
Reported
license unknown · no source
Stellar Flamingo
unknown
32.7µs
2.48×
Reported
license unknown · no source
Stellar Flamingo
unknown
32.9µs
2.50×
Reported
license unknown · no source
Stellar Flamingo
unknown
32.7µs
2.48×
Reported
license unknown · no source
Stellar Flamingo
unknown
32.8µs
2.49×
Reported
license unknown · no source
AKO4ALL
unknown
32.8µs
2.49×
Reported
license unknown · no source
Stellar Flamingo
unknown
33.6µs
2.55×
Reported
license unknown · no source
Stellar Flamingo
unknown
33.8µs
2.57×
Reported
license unknown · no source
Stellar Flamingo
unknown
34.0µs
2.58×
Reported
license unknown · no source
Stellar Flamingo
unknown
34.5µs
2.62×
Reported
license unknown · no source
AKO4ALL
unknown
32.8µs
2.49×
Reported
license unknown · no source
Stellar Flamingo
unknown
37.9µs
2.87×
Reported
license unknown · no source
Stellar Flamingo
unknown
39.8µs
3.02×
Reported
license unknown · no source
Geometric
unknown
41.9µs
3.18×
Reported
license unknown · no source
AKO
unknown
43.2µs
3.28×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
47.6µs
3.62×
Reported
license unknown · no source
Stellar Flamingo
unknown
48.8µs
3.70×
Reported
license unknown · no source
Geometric
unknown
27.9µs
2.12×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
49.6µs
3.77×
Reported
license unknown · no source
doubleAI
unknown
52.2µs
3.96×
Reported
license unknown · no source
KerSor
unknown
85.9µs
6.52×
Reported
license unknown · no source
Geometric
unknown
47.9µs
3.64×
Reported
license unknown · no source
Geometric
unknown
97.5µs
7.40×
Reported
license unknown · no source
Cursor
unknown
165.1µs
12.53×
Reported
license unknown · no source
Geometric
unknown
139.9µs
10.62×
Reported
license unknown · no source
Nimble Penguin
unknown
203.2µs
15.42×
Reported
license unknown · no source
Nimble Penguin
unknown
363.5µs
27.60×
Reported
license unknown · no source
Amir M. Mir | SF Tensor
unknown
845.1µs
64.17×
Reported
license unknown · no source
Nimble Penguin
unknown
1.99ms
151.25×
Reported
license unknown · no source
Geometric
unknown
16.9µs
1.29×
Reported
license unknown · no source
Geometric
unknown
17.8µs
1.35×
Reported
license unknown · no source
RIAC as well
unknown
13.2µs
1.00×
Reported
license unknown · no source
claude-opus-4-1 / cudafeab9bclaude-opus-4-1-20250805
cuda
—
—
Apache-2.0 · source
claude-opus-4-1_triton_a98005claude-opus-4-1-20250805
triton
—
—
Apache-2.0 · source
gemini-2.5-pro_cuda_292432gemini-2.5-pro
cuda
—
—
Apache-2.0 · source
gemini-2.5-pro_triton_dorbxsgemini-2.5-pro
triton
—
—
Apache-2.0 · source
gpt-5 / cuda5eb89cgpt-5-2025-08-07
cuda
46.3µs
3.52×
Reproduction-ready
Apache-2.0 · source
gpt-5 / tritona41cd4gpt-5-2025-08-07
triton
101.2µs
7.68×
Reported
Apache-2.0 · source
gpt-o3 / cuda7cdf4bgpt-o3
cuda
—
—
Apache-2.0 · source
gpt-o3 / triton4c17a1gpt-o3
triton
59.1µs
4.49×
Reported
Apache-2.0 · source
flashinfer / wrapper03f7b0FlashInfer-Bench baselines
python
15.1µs
1.14×
Reported
Apache-2.0 · source

Semantics

Inputs and outputs
q_nopebf16 [batch_size, num_qo_heads, head_dim_ckv]
q_pebf16 [batch_size, num_qo_heads, head_dim_kpe]
ckv_cachebf16 [num_pages, page_size, head_dim_ckv]
kpe_cachebf16 [num_pages, page_size, head_dim_kpe]
kv_indptrint32 [len_indptr]
kv_indicesint32 [num_kv_indices]
sm_scalefp32 scalar
outputbf16 [batch_size, num_qo_heads, head_dim_ckv]
lsefp32 [batch_size, num_qo_heads]
Axes and behavior
num_pagesvariable
page_sizeconstant = 1
batch_sizevariable
len_indptrvariable
head_dim_ckvconstant = 512
head_dim_kpeconstant = 64
num_qo_headsconstant = 16
num_kv_indicesvariable
determinismunspecified
constraintsNo mutation or aliasing
Identity
aliasmla_paged_decode_h16_ckv512_kpe64_ps1modeldeepseek-v3modeldeepseek-r1sha25648be84318f5d…

Also indexed as 018-mla-paged-decode-h16-ckv512-kpe64-ps1 · reviewed as the same computation; every definition's cohorts are shown here, ranked separately.

Sources: NVIDIA SOL-ExecBench (2026-08-18) · FlashInfer-Bench (2025-10-21) · Apache-2.0last observed 2026-08-18How records are decidedJSON