gpt-5 / cudafd76fe
MLA paged prefill causal h16 ckv512 kpe64 ps1 · total_q = 96 · num_pages = 989669 · len_indptr = 2 · num_kv_indices = 98 · bf16 · run 01a02032-596a…
●passedReported evidence
Not re-observed recently.
Primary measurement
139.9µsmean
Rank 2 in its comparison group · source-native comparison · observed 2025-10-16
362.22× reference
Identity
implementationgpt-5_cuda_fd76fe
projectgpt-5-2025-08-07
revisionda915083d4c7c5e61aa3005e3d17ae488e0fc71c
workloadtotal_q = 96 · num_pages = 989669 · len_indptr = 2 · num_kv_indices = 98 · bf16
comparison keysha256:cc2d4c7a3762000a…
sourceFlashInfer-Bench
external idtrace/gpt-5_cuda_fd76fe/9722a8a5-542e-4b96-b962-7ae93eb259ac
sha256:eceed8e4b2714e0e622f203…
Correctness
comparatorflashinfer_bench_eval
max absolute error0.0078125
max relative error0.28268569707870483
resultpassed
Workload
total_q96
num_pages989669
len_indptr2
num_kv_indices98
q_pebf16 [96, 16, 64]
q_nopebf16 [96, 16, 512]
ckv_cachebf16 [989669, 1, 512]
kpe_cachebf16 [989669, 1, 64]
kv_indptrint32 [2]
qo_indptrint32 [2]
kv_indicesint32 [98]
definition comparatorsol_execbench_eval
Measurements
latency · mean139.9 µs
Protocol
harnessflashinfer-bench
timerharness_reported
primaryStatisticmean
correctnessReferencedefinition reference implementation
comparabilityFamilyflashinfer_bench_trace
Environment
gpuNVIDIA B200 (unknown)
cudaToolkit12.8
frameworkpytorch 2.8.0+cu128
Artifacts
No artifacts published with this run.
Replications and notes
No attestations yet.
Community attestations. They never change the evidence level; only a KernelIndex-controlled rerun does.
Add a reproduction or note
Canonical manifest
Show manifestHide manifest
{
"run": {
"kind": "BenchmarkRun",
"spec": {
"status": "passed",
"timing": {
"latencyNs": {
"mean": 139945
},
"primaryStatistic": "mean"
},
"observedAt": "2025-10-16T11:19:45.515Z",
"correctness": {
"comparator": "sol_execbench_eval",
"maximumAbsoluteError": 0.0078125,
"maximumRelativeError": 0.28268569707870483
},
"sourceNative": {
"source": "flashinfer-bench",
"metrics": {
"latency_ms": 0.13994513524434202,
"speedup_factor": 362.22157177582477,
"reference_latency_ms": 50.69114685058594
},
"benchmark": "mla_paged_prefill_causal_h16_ckv512_kpe64_ps1"
},
"protocolDigest": "sha256:7d44d206ad086b4bece35ce31f6531ebb11f07e0d9dcfb74c75be86d3ec02095",
"workloadDigest": "sha256:974d2a224e40a3ad0d59754d4f55edc0c7f3bbf320dac599275fe98138f074c3",
"environmentDigest": "sha256:f4292a8be127b7230cd3995c1d605fb6c644e2c2f94566e1d28cc1be766ca5d5",
"implementationDigest": "sha256:28f7dd8c2d366e4c9dbbcf3ca979748c71ddba7e0dbfd17278c0993d062993db"
},
"metadata": {
"name": "sol-trace-mla-paged-prefill-causal-h16-ckv512-kpe64-ps1-9722a8a5",
"title": "gpt-5_cuda_fd76fe · mla_paged_prefill_causal_h16_ckv512_kpe64_ps1"
},
"apiVersion": "kernelindex.dev/v1alpha1"
},
"protocol": {
"kind": "BenchmarkProtocol",
"spec": {
"harness": {
"name": "flashinfer-bench",
"repository": "https://github.com/flashinfer-ai/flashinfer-bench"
},
"correctness": {
"reference": "definition reference implementation",
"comparator": "flashinfer_bench_eval"
},
"measurement": {
"timer": "harness_reported",
"primaryStatistic": "mean"
},
"comparability": {
"family": "flashinfer_bench_trace"
}
},
"metadata": {
"name": "flashinfer-bench-trace",
"title": "FlashInfer-Bench trace harness"
},
"apiVersion": "kernelindex.dev/v1alpha1"
},
"environment": {
"kind": "ExecutionEnvironment",
"spec": {
"hardware": {
"vendor": "nvidia",
"product": "NVIDIA B200",
"architecture": "unknown"
},
"software": {
"framework": {
"name": "pytorch",
"version": "2.8.0+cu128"
},
"libraries": {
"triton": "3.4.0"
},
"cudaToolkit": "12.8"
}
},
"metadata": {
"name": "sol-trace-nvidia-b200-2-8-0-cu128-3-4-0-12-8",
"title": "NVIDIA B200 (trace-reported software)"
},
"apiVersion": "kernelindex.dev/v1alpha1"
}
}Cite this record (permalink, digest, access date)
Report an issue with this run
Published 2026-08-20 · FlashInfer-Bench · Apache-2.0All results for MLA paged prefill causal h16 ckv512 kpe64 ps1 →JSON