Skip to content
KernelIndex
Search⌘K

FP8 YaRN RoPE embedding

38 eligible runs
rope

FP8-quantized YaRN (Yet another RoPE extensioN method) rotary position embedding computation with dynamic frequency adjustment for extended context (163840 tokens). Computes cos/sin embeddings for rotary position encoding with YaRN interpolation and mscale factors.

Fastest reported
2.00µsmean · 93% of SOL
doubleAILicense unknown · unknown

Reported evidence · last observed 2026-07-10. Reported by source; not independently reproduced. No public source. License unknown.

No public sourceRun detail →

Current records

Workloadsuite of 16 cases · mean latencyhalf_dim = 32 · beta_fast = 32 · beta_slow = 1 · rope_theta = 10000 · scaling_factor = 40 · qk_rope_head_dim = 64 · max_position_embeddings = 163840 · original_max_position_embeddings = 409616 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 · 30 results · last observed 2026-08-24Record history →
#
Implementation
Latency
vs #1
Trust
Observed
1
2.00µs93% SOL
1.00×
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 →
2
2.56µs91% SOL
1.28×
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 →
3
2.56µs91% 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 →
4
2.58µs91% SOL
1.29×
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
2.60µs91% SOL
1.30×
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 →
6
2.60µs91% SOL
1.30×
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 →
7
2.61µs91% SOL
1.31×
Reported · license unknown · no source
2026-08-23

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

no sourcelicense unknownno install recipeRun detail →
8
2.66µs90% SOL
1.33×
Reported · license unknown · no source
2026-08-23

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

no sourcelicense unknownno install recipeRun detail →
9
2.67µs90% SOL
1.34×
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 →
10
2.67µs90% SOL
1.34×
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
2.68µs90% SOL
1.34×
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 →
12
2.68µs90% SOL
1.34×
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 →
13
2.68µs90% SOL
1.34×
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 →
14
2.69µs90% SOL
1.35×
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 →
15
2.75µs90% SOL
1.38×
Reported · license unknown · no source
2026-08-23

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

no sourcelicense unknownno install recipeRun detail →
16
2.81µs90% 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 →
17
2.86µs90% SOL
1.43×
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 →
18
3.06µs89% SOL
1.53×
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 →
19
3.08µs89% SOL
1.54×
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 →
20
3.19µs89% SOL
1.60×
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 →
21
3.23µs88% SOL
1.62×
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 →
22
3.30µs88% SOL
1.65×
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 →
23
3.40µs88% SOL
1.71×
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 →
24
3.41µs88% SOL
1.71×
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 →
25
3.41µs88% SOL
1.71×
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 →
26
3.42µs88% SOL
1.72×
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 →
27
3.60µs87% SOL
1.80×
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 →
28
3.64µs87% SOL
1.82×
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 →
29
4.03µs86% SOL
2.02×
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
4.27µs85% SOL
2.14×
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
2.00µs
1.05×
Reported
license unknown · no source
unknown
2.60µs
1.36×
Reported
license unknown · no source
unknown
2.81µs
1.47×
Reported
license unknown · no source
unknown
1.91µs
1.00×
Reported
license unknown · no source
unknown
2.56µs
1.34×
Reported
license unknown · no source
unknown
2.86µs
1.50×
Reported
license unknown · no source
unknown
3.12µs
1.63×
Reported
license unknown · no source
unknown
3.36µs
1.76×
Reported
license unknown · no source
Show all 38 implementations ›
unknown
3.40µs
1.79×
Reported
license unknown · no source
unknown
3.42µs
1.80×
Reported
license unknown · no source
unknown
3.41µs
1.79×
Reported
license unknown · no source
unknown
3.41µs
1.79×
Reported
license unknown · no source
unknown
4.01µs
2.10×
Reported
license unknown · no source
unknown
4.83µs
2.54×
Reported
license unknown · no source
unknown
2.56µs
1.34×
Reported
license unknown · no source
unknown
2.58µs
1.35×
Reported
license unknown · no source
unknown
2.60µs
1.36×
Reported
license unknown · no source
unknown
2.61µs
1.37×
Reported
license unknown · no source
unknown
2.66µs
1.39×
Reported
license unknown · no source
unknown
2.67µs
1.40×
Reported
license unknown · no source
unknown
2.67µs
1.40×
Reported
license unknown · no source
unknown
2.68µs
1.41×
Reported
license unknown · no source
unknown
2.68µs
1.41×
Reported
license unknown · no source
unknown
2.68µs
1.40×
Reported
license unknown · no source
unknown
2.69µs
1.41×
Reported
license unknown · no source
unknown
2.75µs
1.44×
Reported
license unknown · no source
unknown
3.08µs
1.61×
Reported
license unknown · no source
unknown
3.06µs
1.60×
Reported
license unknown · no source
unknown
3.09µs
1.62×
Reported
license unknown · no source
unknown
3.19µs
1.67×
Reported
license unknown · no source
unknown
3.23µs
1.69×
Reported
license unknown · no source
unknown
3.30µs
1.73×
Reported
license unknown · no source
unknown
3.64µs
1.91×
Reported
license unknown · no source
unknown
3.60µs
1.89×
Reported
license unknown · no source
unknown
4.03µs
2.11×
Reported
license unknown · no source
unknown
4.14µs
2.17×
Reported
license unknown · no source
unknown
4.27µs
2.24×
Reported
license unknown · no source
unknown
4.34µs
2.28×
Reported
license unknown · no source

Semantics

Inputs and outputs
position_idsint64 [seq_len]
mscalefp32 scalar
mscale_all_dimfp32 scalar
cos_embbf16 [seq_len, qk_rope_head_dim]
sin_embbf16 [seq_len, qk_rope_head_dim]
Axes and behavior
seq_lenvariable
half_dimhalf_dim = qk_rope_head_dim // 2
beta_fastconstant = 32
beta_slowconstant = 1
rope_thetaconstant = 10000
scaling_factorconstant = 40
qk_rope_head_dimconstant = 64
max_position_embeddingsconstant = 163840
original_max_position_embeddingsconstant = 4096
determinismunspecified
constraintsNo mutation or aliasing
Identity
alias014_fp8_yarn_rope_embeddingmodeldeepseek-v3-1-terminussha256b19798339b0f…
Sources: NVIDIA SOL-ExecBench (2026-08-24)last observed 2026-08-24How records are decidedJSON