Compare runs
cohort sha256:2a26200dc1c84d54… · source-native profile
ranking-v1
no ranks: not comparable
Comparison identity
operation
gqa-paged-decode-h32-kv4-d128-ps1
workload
num_pages = 61535 · batch_size = 64 · len_indptr = 65 · num_kv_indices = 61519 · bf16 · a819523e
protocol
flashinfer-bench · 7d44d206
environment
NVIDIA B200 · af7cb68a
correctness policy
64fb869e
metric
latency mean (ns)
Context · 10 fields›
architecture
unknown
CUDA
12.8
driver
unknown
framework
pytorch 2.9.1+cu128
samples
unknown
evidence
reported
license
Apache-2.0
source
available
status
passed
observed
2026-03-28
No winner: these runs didn't measure the same thing. The rows above show what differs. Why comparable?
Add run ›
Export ›
Markdown CSV JSON
gpt-5_triton_f88811 · sha256:584288007d2f6… · observed 2026-03-28