Skip to content
KernelIndex
Search⌘K

gemini-2.5-pro

Competition author · flashinfer-ai/flashinfer-trace · Apache-2.0
Records held
95as of 2026-10-01 12:02 UTC
Entries
29
Runs
632
Is this you? Claim this project ›

A claim grants attribution and metadata maintenance, never the right to edit evidence. Evidence claims are reviewed by a maintainer.

Current records

Operation · workload
History
Record
Implementation
Hardware
Since
RMSNorm h7168bf16 · [7168] · batch 539
10.2µs
0.0% faster · was 10.2 µs
B200
2025-10-16
RMSNorm h7168bf16 · [7168] · batch 14521
73.7µs
12.3% faster · was 84.0 µs
B200
2025-10-16
RMSNorm h7168bf16 · [7168] · batch 11949
63.2µs
11.6% faster · was 71.6 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 14418
41.4µs
40.7% faster · was 69.7 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 11832
35.7µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 64
7.90µs
22.4% faster · was 10.2 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 16
7.27µs
27.8% faster · was 10.1 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 10827
33.9µs
36.2% faster · was 53.1 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 8804
28.8µs
36.0% faster · was 45.0 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 63
7.62µs
46.6% faster · was 14.3 µs
B200
2025-10-16
Show all 40 records ›
RMSNorm h4096bf16 · [4096] · batch 14509
41.7µs
40.4% faster · was 70.0 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 15
7.57µs
47.0% faster · was 14.3 µs
B200
2025-10-16
RMSNorm h512bf16 · [512] · batch 32
6.18µs
B200
2025-10-16
RMSNorm h512bf16 · [512] · batch 11949
13.5µs
13.1% faster · was 15.5 µs
B200
2025-10-16
RMSNorm h512bf16 · [512] · batch 64
6.19µs
34.3% faster · was 9.41 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 32
6.26µs
15.3% faster · was 7.39 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 18
6.51µs
12.6% faster · was 7.45 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 14521
20.6µs
52.1% faster · was 43.0 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 11949
18.5µs
49.9% faster · was 36.9 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 64
6.29µs
3.6% faster · was 6.52 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 6
6.27µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 1
6.21µs
56.6% faster · was 14.3 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 34
6.17µs
49.9% faster · was 12.3 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 79
6.22µs
49.5% faster · was 12.3 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 16254
31.1µs
24.6% faster · was 41.2 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 12383
25.7µs
22.3% faster · was 33.1 µs
B200
2025-10-16
RMSNorm h2048bf16 · [2048] · batch 64
6.18µs
18.3% faster · was 7.56 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 7
7.30µs
28.5% faster · was 10.2 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 1
7.40µs
27.3% faster · was 10.2 µs
B200
2025-10-16
RMSNorm h4096bf16 · [4096] · batch 34
7.46µs
26.0% faster · was 10.1 µs
B200
2025-10-16
Fused add RMSNorm h7168bf16 · [7168] · batch 11949
82.3µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 1088
6.20µs
0.4% faster · was 6.23 µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 520128
273.7µs
0.3% faster · was 274.4 µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 65016
39.2µs
4.1% faster · was 40.9 µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 396256
209.8µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 49532
32.0µs
2.1% faster · was 32.7 µs
B200
2025-10-16
RMSNorm h128bf16 · [128] · batch 2048
7.29µs
2.4% faster · was 7.47 µs
B200
2025-10-16
RMSNorm h1536bf16 · [1536] · batch 7
6.32µs
24.2% faster · was 8.34 µs
B200
2025-10-16
Fused add RMSNorm h4096bf16 · [4096] · batch 8804
43.1µs
19.7% faster · was 53.7 µs
B200
2025-10-16
Fused add RMSNorm h4096bf16 · [4096] · batch 63
7.42µs
20.3% faster · was 9.31 µs
B200
2025-10-16

55 more records · Open in the records ledger →

Measured entries

Implementation
Operation
Runtime
Best latency
Evidence
Availability
triton
6.17µs
Reported
Apache-2.0 · source
triton
6.18µs
Reported
Apache-2.0 · source
triton
6.18µs
Reported
Apache-2.0 · source
cuda
6.21µs
Reproduction-ready
Apache-2.0 · source
triton
6.26µs
Reported
Apache-2.0 · source
cuda
6.26µs
Reproduction-ready
Apache-2.0 · source
triton
6.52µs
Reported
Apache-2.0 · source
triton
6.64µs
Reported
Apache-2.0 · source
cuda
7.27µs
Reproduction-ready
Apache-2.0 · source
triton
8.19µs
Reported
Apache-2.0 · source
Show all 29 entries ›
triton
8.20µs
Reported
Apache-2.0 · source
triton
9.30µs
Reported
Apache-2.0 · source
cuda
10.3µs
Reproduction-ready
Apache-2.0 · source
cuda
15.8µs
Reproduction-ready
Apache-2.0 · source
triton
16.8µs
Reported
Apache-2.0 · source
cuda
18.3µs
Reproduction-ready
Apache-2.0 · source
triton
24.6µs
Reported
Apache-2.0 · source
triton
25.0µs
Reported
Apache-2.0 · source
triton
39.0µs
Reported
Apache-2.0 · source
triton
43.0µs
Reported
Apache-2.0 · source
triton
45.1µs
Reported
Apache-2.0 · source
triton
64.9µs
Reported
Apache-2.0 · source
triton
67.7µs
Reported
Apache-2.0 · source
triton
68.1µs
Reported
Apache-2.0 · source
triton
97.6µs
Reported
Apache-2.0 · source
triton
128.2µs
Reported
Apache-2.0 · source
triton
151.7µs
Reported
Apache-2.0 · source
triton
162.4µs
Reported
Apache-2.0 · source
cuda
409.5µs
Reproduction-ready
Apache-2.0 · source

Record activity

No record transitions on this GPU in the trailing year.

Sources: FlashInfer-Bench (2025-10-21) · Apache-2.0last observed 2025-10-21JSON