cuda_nvfp4_grouped_naive_g4_n2048_k2048
Grouped GEMM NVFP4 m contiguous g4 n2048 k2048 · m = 512 · fp32 · run 01a00b1d-14e4…
●passedReproduction-ready evidence
Primary measurement
19.7msmean
Rank 1 in its comparison group · source-native comparison · observed 2026-06-06
0.11× reference
Identity
implementationcuda_nvfp4_grouped_naive_g4_n2048_k2048
projectFlashInfer-Bench baselines
revisionda915083d4c7c5e61aa3005e3d17ae488e0fc71c
workloadm = 512 · fp32
comparison keysha256:2199848e5695ee73…
sourceFlashInfer-Bench
external idtrace/cuda_nvfp4_grouped_naive_g4_n2048_k2048/a53a7570-79be-4e19-b64f-a4d2c42ca985
sha256:7dd7ed364c73b7809447ecc…
Correctness
comparatorflashinfer_bench_eval
max absolute error0
max relative error0
resultpassed
Workload
m512
a_fp4int8 [512, 1024]
alphafp32 [4]
b_fp4int8 [4, 2048, 1024]
a_scaleint8 [512, 128]
b_scaleint8 [4, 2048, 128]
m_indptrint32 [5]
definition comparatorsol_execbench_eval
Measurements
latency · mean19.7 ms
Protocol
harnessflashinfer-bench
timerharness_reported
primaryStatisticmean
correctnessReferencedefinition reference implementation
comparabilityFamilyflashinfer_bench_trace
Environment
gpuNVIDIA B200 (unknown)
cudaToolkit13.0
frameworkpytorch 2.11.0+cu130
Artifacts
logssha256:8a42c797d4b34712… · text/plain · unavailable
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": 19682148
},
"primaryStatistic": "mean"
},
"evidence": {
"logs": {
"uri": "kernelindex:artifact/sha256:8a42c797d4b34712c43e8e198fc0f346aa83d9b9e6b18866e8db980cd751e0dc",
"digest": "sha256:8a42c797d4b34712c43e8e198fc0f346aa83d9b9e6b18866e8db980cd751e0dc"
}
},
"observedAt": "2026-06-06T22:07:53.817Z",
"correctness": {
"comparator": "sol_execbench_eval",
"maximumAbsoluteError": 0,
"maximumRelativeError": 0
},
"sourceNative": {
"source": "flashinfer-bench",
"metrics": {
"latency_ms": 19.682148,
"speedup_factor": 0.10978241128288774,
"reference_latency_ms": 2.1607536666666665
},
"benchmark": "grouped_gemm_nvfp4_m_contiguous_g4_n2048_k2048"
},
"protocolDigest": "sha256:7d44d206ad086b4bece35ce31f6531ebb11f07e0d9dcfb74c75be86d3ec02095",
"workloadDigest": "sha256:8103ee699d3eddb7c8198fd19d1a450ed7ad0943d4ecb58317f7d3c752e50e28",
"environmentDigest": "sha256:83c9b6e78d493927f18878d57448d9a0d2221a70840b562455b2665a2897ba19",
"implementationDigest": "sha256:6f52d7622602ed94cd3c9f5f2dd3e8ab27f508b76b8e76c9108e293e295c469a"
},
"metadata": {
"name": "sol-trace-grouped-gemm-nvfp4-m-contiguous-g4-n2048-k2048-a53a7570",
"title": "cuda_nvfp4_grouped_naive_g4_n2048_k2048 · grouped_gemm_nvfp4_m_contiguous_g4_n2048_k2048"
},
"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": "2.11.0+cu130"
},
"libraries": {
"triton": "3.6.0"
},
"cudaToolkit": "13.0"
}
},
"metadata": {
"name": "sol-trace-nvidia-b200-2-11-0-cu130-3-6-0-13-0",
"title": "NVIDIA B200 (trace-reported software)"
},
"apiVersion": "kernelindex.dev/v1alpha1"
}
}Cite this record (permalink, digest, access date)
Report an issue with this run
Published 2026-08-16 · FlashInfer-Bench · Apache-2.0All results for Grouped GEMM NVFP4 m contiguous g4 n2048 k2048 →JSON