LOADING DATASET
Opening the index…
Resolving the selected comparison slice.
Resolving the selected comparison slice.
Find the fastest verified inference hardware for your workload.
Within a pinned MLPerf slice, after source and artifact checks. These are submitted systems: accelerator count, topology and software matter. This index is not independent hardware certification.
Output throughput
tokens / second ↑
4 matching entries · ties share competition rank · each variant’s best submitted system
| Rank | Accelerator / vendor | Submitted system | Count | Throughput | Evidence |
|---|---|---|---|---|---|
| 1 | GB300 NVIDIA · GB300 · 288 GB | NVIDIA GB300-NVL72_GB300-288GB_aarch64x288_TRT | 288 | 1,555,110 | Sources ↗run_1/mlperf_log_summary.txtSHA-256 0e24cd10b61d…systems/GB300-NVL72_GB300-288GB_aarch64x288_TRT.jsonSHA-256 6b1dbc79026a…run_1/mlperf_log_detail.txtSHA-256 1201de13995c…accuracy/accuracy.txtSHA-256 6edcead7c8b2…accuracy/mlperf_log_detail.txtSHA-256 83a81d914156…TEST06/verify_accuracy.txtSHA-256 4996c4b15e71… |
| 2 | GB200 NVIDIA · GB200 · 186 GB | NVIDIA GB200-NVL72_GB200-186GB_aarch64x72_TRT | 72 | 336,106 | Sources ↗run_1/mlperf_log_summary.txtSHA-256 cb17d749df1a…systems/GB200-NVL72_GB200-186GB_aarch64x72_TRT.jsonSHA-256 e11f427b7e1b…run_1/mlperf_log_detail.txtSHA-256 5142e5b68e8c…accuracy/accuracy.txtSHA-256 f2613b250235…accuracy/mlperf_log_detail.txtSHA-256 db348a521d31…TEST06/verify_accuracy.txtSHA-256 4996c4b15e71… |
| 3 | B300 NVIDIA · SXM · 270 GB | Nebius nebius_b300_n1 | 8 | 60,413.4 | |
| 4 | B200 NVIDIA · SXM · 180 GB | Nebius nebius_b200_n1 | 8 | 51,692.9 |