LLM serving (MLPerf datacenter) benchmarks
MLPerf Inference datacenter LLM serving (server and offline scenarios). Kept separate from edge LLM generation and not used in edge rankings.
605 published rows. Each table compares only results with the same model-size class and batch size. Search this workload · JSON.
Fastest in the best-covered like-for-like group (LLM >40B, batch unknown): Crusoe Cloud 64x8 MI355X (8x AMD Instinct MI355X 288GB HBM3e) at 5749440 tok/s A 2026-09-16 (published). 3 groups, 118 platforms.
Show data — 3 like-for-like tables
LLM >40B · batch unknown (93 platforms)
✓ Like-for-like: model-size class LLM >40B, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Crusoe Cloud 64x8 MI355X (8x AMD Instinct MI355X 288GB HBM3e) | 5749440 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b mxfp4 · vLLM 0.22.1 (ROCm 7.2.2), PyTorch 2.10 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| CoreWeave GB300 NVL72 (4x NVIDIA GB300) | 1197710 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Nebius GB300 NVL72 (4x NVIDIA GB300) | 1186750 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB300 NVL72 (4x NVIDIA GB300) | 1126850 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| AMD MI355X Custom Cluster (8x AMD Instinct MI355X 288GB HBM3e (x87)) | 1042110 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · PyTorch 2.9.1+git8907517, ROCm 7.0.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 9x8xMI355X_2xEPYC_9575F (8x AMD Instinct MI355X 288GB HBM3e) | 1039900 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b mxfp4 · vLLM 0.22.1.dev0, PyTorch 2.10.0, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AMD MI355X Custom Cluster (8x AMD Instinct MI355X 288GB HBM3e (x94)) | 1031070 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · PyTorch 2.9.1+git8907517, ROCm 7.0.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| CoreWeave GB200 NVL72 (4x NVIDIA GB200) | 911566 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB200 NVL72 (4x NVIDIA GB200) | 899218 tok/s throughput_server · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Heterogeneous Four-Cluster (8x AMD Instinct MI300X 192GB HBM3 (x2x8) and AMD Instinct MI355X 288GB HBM3e (x2x8)) | 285454 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · ROCm · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| MangoBoost 2x8 MI355X (8x AMD Instinct MI355X 288GB HBM3e) | 228488 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · ROCm 7.2.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco_UCS_B300x16_G200_4x2_RO (8x NVIDIA B300-SXM-270GB) | 227570 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS 16xMI350X (8x AMD Instinct MI350X 288GB HBM3e) | 159932 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell_MangoBoost Heterogeneous Cluster (8x AMD Instinct MI300X 192GB HBM3 (x8) and AMD Instinct MI325X 256GB HBM3e (x8) and AMD Instinct MI355X 288GB HBM3e (x8)) | 151843 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · ROCm · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Orrick Optimized 6x NVIDIA B200 L2 (6x NVIDIA B200-SXM-180GB) | 130928 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 nvfp4 · TensorRT-LLM 1.3.0rc20 · LLM >40B | B | unknown power n/a no power published | MLPerf Inference v6.1, open division, data-centre suite 2026-09-16 (published) |
| smci355-ccs-aus-m13-13 (8x AMD Instinct MI355X 288GB HBM3e) | 121818 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · PyTorch 2.10.0+git8514f05, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G4826 (8x AMD Instinct MI355X 288GB HBM3e) | 118042 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell XE9785L_8xMI355X (8x AMD Instinct MI355X 288GB HBM3e) | 117804 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB HBM3e) | 117355 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AS-4126GS-NMR-LCC (8x AMD Instinct MI355X 288GB HBM3e) | 117005 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| PowerEdge XE9785L (8x AMD Instinct MI355X 288GB HBM3e) | 116953 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · ROCm · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB) | 115730 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.13, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 8xMI355X_2xEPYC_9575F (8x AMD Instinct MI355X 288GB HBM3e) | 115567 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| CoreWeave B300 SXM (8x NVIDIA B300-SXM-270GB) | 115530 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML:mlperf-inf-mm-q3vl-v6.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| PowerEdge XE9785 (8x AMD Instinct MI355X 288GB HBM3e) | 115294 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · ROCm · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-AD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 115034 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-ZD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 114826 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE Compute XD690 (8x NVIDIA B300-SXM-270GB) | 114569 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| QuantaGrid D75H-10U (8x NVIDIA B300-SXM-270GB) | 113985 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| XA NB3I-E12 (8x NVIDIA B300-SXM-270GB) | 113455 tok/s throughput_server · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell XE9785L (8x NVIDIA B300-SXM-270GB) | 113101 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA DGX B300 (8x NVIDIA B300-SXM-270GB) | 112954 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| SYS-822GS-NB3RT (8x NVIDIA B300-SXM-270GB) | 111247 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AS-8126GS-NB3RT (8x NVIDIA B300-SXM-270GB) | 110802 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Inventec P5800G7 (8x NVIDIA B300-SXM-270GB) | 109933 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Inventec P9000G7 (8x NVIDIA B300-SXM-270GB) | 109755 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-SD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 109343 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell XE9780 (8x NVIDIA B300-SXM-270GB) | 109257 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, Dell Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentML: · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Lenovo ThinkSystem SR680a V4 (8x NVIDIA B300-SXM-270GB) | 109141 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE9780L (8x NVIDIA B300-SXM-270GB) | 108669 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Nebius B300 n1 (8x NVIDIA B300-SXM-270GB) | 108529 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute XD685 (8x NVIDIA B200-SXM-180GB) | 103954 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.11, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| smci355-ccs-aus-m09-17 (8x AMD Instinct MI355X 288GB HBM3e) | 103480 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · PyTorch 2.9.1+git8907517, ROCm 7.0.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| CoreWeave B200 SXM (8x NVIDIA B200-SXM-180GB) | 102703 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| SYS-822GS-NBRT (8x NVIDIA B200-SXM-180GB) | 101595 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| NVIDIA GB200 NVL72 by HPE (4x NVIDIA GB200) | 101453 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| QuantaGrid D74A-7U (8x AMD Instinct MI355X 288GB HBM3e) | 100754 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TODO · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| G893-ZX1-AAX4 (8x AMD Instinct MI355X 288GB HBM3e) | 100310 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · PyTorch 2.9.1+git8907517, ROCm 7.0.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Inventec P9000AG7 (8x NVIDIA B200-SXM-180GB) | 100291 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.11, CUDA 12.9 · LLM >40B | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE9785L (8x AMD Instinct MI355X 288GB HBM3e) | 100008 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · vLLM 0.14.0rc3.dev30+gb026cf14e.d20260212.rocm700, Pytorch 2.9.1+git8907517, ROCm 7.0.0.70000-38~22.04 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| BM.GPU.MI355X.8 (8x AMD Instinct MI355X 288GB HBM3e) | 99874 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR (8x NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e) | 99684 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp8,fp4 · vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G8825Z5 (8x AMD Instinct MI350X 288GB HBM3e) | 97020 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · Pytorch 2.10, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| LLM-D v0.5.0,Openshift 4.20.12,NVIDIA 8xB200-SXM-180GB (8x NVIDIA B200-SXM-180GB) | 93071 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b mxfp4 · LLM-D v0.5.0 ,vLLM 0.14.1, RHEL 9.6 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Nebius B200 n1 (8x NVIDIA B200-SXM-180GB) | 89856 tok/s throughput_server · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| PowerEdge XE9785 (8x AMD Instinct MI355X 288GB HBM3e (Power Cap 1000 W)) | 81871 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · ROCm · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| AS-8126GS-TNMR,vLLM (8x AMD Instinct MI350X 288GB HBM3e) | 80478 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · vLLM 0.14.0, RHEL 9.6 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS 8xMI350X (8x AMD Instinct MI350X 288GB HBM3e) | 78937 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Lambda 4x GB300-288GB (4x NVIDIA GB300) | 65512 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b mxfp4 · TensorRT-LLM 1.3.0rc14 (GPT-OSS-120B); vLLM (CentML) 0.20.2rc1.dev2091+g2e1040b7e + NVIDIA Dynamo (Qwen3-VL-235B); CUDA · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AS-8126GS-TNMR,vLLM (8x AMD Instinct MI350X) | 64293 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · · LLM >40B | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| OneW GB300 NVL72 (4x NVIDIA GB300) | 63991 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| SMC Lab OpenShift GB200-NVL4 4xGB200-186GB (4x NVIDIA GB200) | 61206 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · vLLM 0.24.0, Red Hat OpenShift 4.21 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| XE7745_8xMI350P_2xEPYC_9455 (8x AMD Instinct MI350P) | 48971 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 36711 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14.1.48, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| MangoBoost Mi325X (8x AMD Instinct MI325X 256GB HBM3e) | 33880 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · ROCm 6.10.5 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE7740 (8x NVIDIA H200-NVL-141GB) | 32004 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · TensorRT 10.13, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 8xMI300X_2xEPYC_9534 (8x AMD Instinct MI300X 192GB HBM3) | 30198 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp8 · vLLM 0.24.0 (stock vllm/vllm-openai-rocm:v0.24.0), ROCm 7.2.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| MiTAC_G4520G6_RTX_PRO_6000_PCIE_96GBx8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 30034 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.2 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 29908 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14.1.48, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS C845A M8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 29904 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell PowerEdge XE7740 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 29797 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| LLM-D v0.5.0,OpenShift 4.21,IBM cloud gx3d-160x1792x8h200,NVIDIA 8xH200-SXM-141GB (8x NVIDIA H200-SXM-141GB) | 28680 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b mxfp4 · LLM-D v0.5.0 ,OpenShift 4.21, RHEL 9.6 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 28039 tok/s throughput_server · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 27730 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| MangoBoost Mi300X (8x AMD Instinct MI300X 192GB HBM3) | 27599 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · ROCm 6.10.5 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| PRIMERGY_CDI_RTX_PRO_6000_PCIE_96GBx8) (8x NVIDIA RTX PRO 6000 Blackwell) | 27341 tok/s throughput_server · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.16, CUDA 13.2, cuDNN 9.21 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Lenovo ThinkSystem SR675i V3 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 27160 tok/s throughput_server · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE9680 (8x AMD Instinct MI300X 192GB HBM3) | 27073 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp8 · vLLM 0.9.0.2.dev108+g71faa1880.d20260213, Pytorch 2.7.0+gitf717b2a, ROCm 6.4.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE7745 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 26830 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| HPE Cray XD670 (8x NVIDIA H200-SXM-141GB) | 25940 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp16 · vLLM v0.15.0-dev · LLM >40B | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| NVIDIA DGX B200 (8x NVIDIA B200-SXM-180GB) | 16167 tok/s throughput_server · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| Nebius RTX PRO 6000 n1 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 15739 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Nebius GB300 NVL72 (1x NVIDIA GB300) | 15372 tok/s throughput_offline · b=unknown | MLPerf gpt-oss-120b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, Driver 580.105.08, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS X210c M8 - X580P PCIe Node (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 14861 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| B300-SXM-270GBx1_TRT (1x NVIDIA B300-SXM-270GB) | 14637 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS X215c M8 - X580P PCIe Node (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 13530 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.11, CUDA 12.9 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 13362 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.14.1.48, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 6862 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 fp4 · TensorRT 10.11.0.33, CUDA 13.0 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE7745 (8x NVIDIA RTX PRO 4500 Blackwell) | 4374 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 fp4 · TensorRT 10.14, CUDA 13.3 · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 1-node-4x-BMG-Pro-B60-Dual (4x MS-Intel Arc Pro B60 Dual 48G Turbo) | 3271 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 INT8 · PyTorch vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 1-node-4x-BMG-B70 (4x Intel(R) Arc(TM) Pro B70) | 2465 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99.9 INT8 · PyTorch · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 1-node-4x-BMG-B70 (4x Intel(R) Arc Pro(R) B70) | 2459 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 INT8 · PyTorch vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 1-node-4x-BMG-B60 (4x Intel(R) Arc Pro(R) B60) | 1698 tok/s throughput_offline · b=unknown | MLPerf llama2-70b-99 INT8 · PyTorch vLLM · LLM >40B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
LLM 5–14B · batch unknown (48 platforms)
✓ Like-for-like: model-size class LLM 5–14B, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Orrick Optimized 6x NVIDIA B200 (6x NVIDIA B200-SXM-180GB) | 274927 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b nvfp4 · TensorRT-LLM 1.3.0rc20 · LLM 5–14B | B | unknown power n/a no power published | MLPerf Inference v6.1, open division, data-centre suite 2026-09-16 (published) |
| Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB) | 173950 tok/s throughput_server · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| XA NB3I-E12 (8x NVIDIA B300-SXM-270GB) | 166745 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| NVIDIA DGX B300 (8x NVIDIA B300-SXM-270GB) | 165432 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS H200 x8 + MI350X x8 G200 2x2 TOR (8x NVIDIA H200-SXM-141GB + AMD Instinct MI350X 288GB HBM3e) | 163292 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8,fp4 · vLLM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA DGX B200 (8x NVIDIA B200-SXM-180GB) | 160403 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| G4826 (8x AMD Instinct MI355X 288GB HBM3e) | 159278 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0+git8514f05, ROCm 7.2.53211 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AS-4126GS-NMR-LCC (8x AMD Instinct MI355X 288GB HBM3e) | 159060 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.4 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell XE9785L_8xMI355X (8x AMD Instinct MI355X 288GB HBM3e) | 158458 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute XD685 (8x AMD Instinct MI355X 288GB HBM3e) | 157694 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute XD685 (8x NVIDIA B200-SXM-180GB) | 152625 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| G894-SD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 150707 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 8xMI355X_2xEPYC_9575F (8x AMD Instinct MI355X 288GB HBM3e) | 148594 tok/s throughput_server · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Inventec P9000AG7 (8x NVIDIA B200-SXM-180GB) | 142508 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11, CUDA 12.9 · LLM 5–14B | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| G8825Z5 (8x AMD Instinct MI350X 288GB HBM3e) | 129943 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · Pytorch 2.10, ROCm 7.2.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| XE7745_8xMI350P_2xEPYC_9455 (8x AMD Instinct MI350P) | 73643 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · vLLM 0.22.0, PyTorch 2.10.0, ROCm 7.2.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute DL380a Gen12 (10x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 60680 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14.1.48, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE7740 (8x NVIDIA H200-NVL-141GB) | 53194 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 · TensorRT 10.13, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| HPE ProLiant Compute DL380a Gen12 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 50965 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.16.1.11, CUDA 13.2 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS C845A M8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 50454 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell PowerEdge XE7745 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 49036 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| ESC8000A_E13P_RTX_PRO_6000_PCIE_96GBx8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 48791 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| ESC8000_E12P_RTX_PRO_6000_PCIE_96GBx8 (8x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 48614 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA H200-NVL-141GB) | 25732 tok/s throughput_server · b=unknown | MLPerf llama3.1-8b fp8 · TensorRT 10.11, CUDA 12.9 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS X215c M8 - X580P PCIe Node (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 24914 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| ManyCoreSoft DeepGadget dg5W (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 24512 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14.1.48, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant DL385 Gen11 (4x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 23266 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11.0.33, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| B300-SXM-270GBx1_TRT (1x NVIDIA B300-SXM-270GB) | 21266 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco UCS C845A M8 (8x NVIDIA RTX PRO 4500 Blackwell) | 18801 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant DL385 Gen11 (2x NVIDIA RTX PRO 6000 Blackwell Server Edition) | 12433 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11.0.33, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS C240 M8 - (5x NVIDIA RTX PRO 4500 Blackwell) | 12018 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE ProLiant Compute DL380 Gen12 (5x NVIDIA RTX PRO 4500 Blackwell) | 11665 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11.0.33, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS X210 M8 - X580P PCIe Node (4x NVIDIA RTX PRO 4500 Blackwell) | 9519 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.14, CUDA 13.1 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| H200-1xGPU (1x NVIDIA H200-SXM-141GB) | 7797 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp8 (static per-tensor PTQ, NVIDIA ModelOpt) · TensorRT-LLM 1.0.0 (FP8 W8A8 engine built with trtllm-build, served by trtllm-serve); MLCommons endpoints (LoadGen++) cl · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 1-node-4x-BMG-Pro-B60-Dual (4x MS-Intel Arc Pro B60 Dual 48G Turbo) | 6762 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b INT8 · PyTorch vLLM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 1-node-4x-BMG-B70 (4x Intel(R) Arc Pro(R) B70) | 4617 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b INT8 · PyTorch vLLM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| 1-node-4x-BMG-B70 (4x Intel(R) Arc(TM) Pro B70) | 4613 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 1-node-4x-BMG-B60 (4x Intel(R) Arc Pro(R) B60) | 3347 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b INT8 · PyTorch vLLM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| HPE ProLiant DL145 Gen11 (1x NVIDIA RTX PRO 4500 Blackwell) | 2341 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b fp4 · TensorRT 10.11.0.33, CUDA 13.0 · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco system (Intel Xeon 6788P, CPU only) CPU-only | 2069 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Supermicro system (Intel Xeon 6788P, CPU only) CPU-only | 2063 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Intel system (Intel Xeon 6980P, CPU only) CPU-only | 1917 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| 1-node-4x-BMG-B50 (4x Intel(R) Arc Pro(R) B50) | 1722 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b INT8 · PyTorch vLLM · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| RedHat_Intel system (Intel Xeon 6972P, CPU only) CPU-only | 1507 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell system (Intel Xeon 6787P, CPU only) CPU-only | 1228 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Quanta_Cloud_Technology system (Intel Xeon 6788P, CPU only) CPU-only | 1171 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Oracle system (Intel Xeon 6987P-C, CPU only) CPU-only | 853 tok/s throughput_offline · b=unknown | MLPerf llama3_1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Cisco system (Intel Xeon 6787P, CPU only) CPU-only | 806 tok/s throughput_offline · b=unknown | MLPerf llama3.1-8b INT8 · PyTorch · LLM 5–14B | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
unknown · batch unknown (32 platforms)
✓ Like-for-like: model-size class unknown, batch unknown (size or batch not stated by source — unverified). Model, precision and stack may still differ.
| Platform | Result | Model · precision · stack | Grade | Power | Source |
|---|---|---|---|---|---|
| Crusoe Cloud 64x8 MI355X (8x AMD Instinct MI355X 288GB HBM3e) | 2901950 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · SGLang 0.5.15.post1 (ROCm 7.2.0), PyTorch 2.x · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB300 NVL72 (4x NVIDIA GB300) | 2705130 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Azure GB200 NVL72 (4x NVIDIA GB200) | 2091190 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA Vera Rubin NVL72 (4x NVIDIA Vera Rubin) | 1183327 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Nebius GB300 NVL72 (4x NVIDIA GB300) | 689961 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Azure GB300 NVL72 (4x NVIDIA GB300) | 661143 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| CoreWeave GB300 NVL72 (4x NVIDIA GB300) | 627653 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | B | unknown power n/a no power published | MLPerf Inference v6.0, open division, data-centre suite 2026-04-01 (published) |
| Nebius VR200 NVL72 (4x NVIDIA VR200) | 591368 tok/s throughput_server · b=unknown | MLPerf deepseek-r1 fp4 · TRTLLM-rubin, vLLM-CentML:mlperf-inf-mm-q3vl-v6.1, Dynamo · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| CoreWeave GB200 NVL72 (4x NVIDIA GB200) | 529991 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB200 NVL72 (4x NVIDIA GB200) | 510781 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · PyTorch 2.10.0a0+b4e4ee81d3.nv25.12, TensorRT 10.14.1.48, CUDA Toolkit 13.1, cuDNN 9.17.0, TensorRT-LLM 1.3.0rc0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA GB200 NVL72 by HPE (4x NVIDIA GB200) | 336125 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0 · unknown | B | unknown power n/a no power published | MLPerf Inference v6.1, open division, data-centre suite 2026-09-16 (published) |
| Cisco_UCS_B300x16_G200_4x2_RO (8x NVIDIA B300-SXM-270GB) | 137067 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14.1.48; TensorRT-LLM 1.3.0rc0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| HPE Compute XD690 (8x NVIDIA B300-SXM-270GB) | 136854 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.2 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-AD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 72445 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-ZD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 72327 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| NVIDIA DGX B300 (8x NVIDIA B300-SXM-270GB) | 70326 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Cisco UCS C880A M8 (8x NVIDIA B300-SXM-270GB) | 70325 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Inventec P5800G7 (8x NVIDIA B300-SXM-270GB) | 70270 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| XA NB3I-E12 (8x NVIDIA B300-SXM-270GB) | 69847 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| QuantaGrid D75H-10U (8x NVIDIA B300-SXM-270GB) | 69842 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| AS-8126GS-NB3RT (8x NVIDIA B300-SXM-270GB) | 69830 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Nebius B300 n1 (8x NVIDIA B300-SXM-270GB) | 69655 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| G894-SD3-AAX7 (8x NVIDIA B300-SXM-270GB) | 69617 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| SYS-822GS-NB3RT (8x NVIDIA B300-SXM-270GB) | 69104 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Inventec P9000G7 (8x NVIDIA B300-SXM-270GB) | 69062 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Dell PowerEdge XE9780 (8x NVIDIA B300-SXM-270GB) | 69021 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE9780L (8x NVIDIA B300-SXM-270GB) | 66288 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| ScitiX B200 (8x NVIDIA B200-SXM-180GB) | 60411 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT-LLM 1.3.0rc0, CUDA 13.1 · unknown | B | unknown power n/a no power published | MLPerf Inference v6.1, open division, data-centre suite 2026-09-16 (published) |
| CoreWeave B200 SXM (8x NVIDIA B200-SXM-180GB) | 58913 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.1, closed division, data-centre suite 2026-09-16 (published) |
| Nebius B200 n1 (8x NVIDIA B200-SXM-180GB) | 58582 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1, cuDNN 9.17, TensorRT-LLM feat/1.2-mlpinf, NVIDIA Dynamo mlperf-v6.0-dynamo-v0.8.0, vLLM CentM · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| Dell PowerEdge XE9780 (8x NVIDIA B200-SXM-180GB) | 58235 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
| NVIDIA DGX B200 (8x NVIDIA B200-SXM-180GB) | 57117 tok/s throughput_offline · b=unknown | MLPerf deepseek-r1 fp4 · TensorRT 10.14, CUDA 13.1 · unknown | A | unknown power n/a no power published | MLPerf Inference v6.0, closed division, data-centre suite 2026-04-01 (published) |
How to read these numbers
Every number is labelled. Confidence: A MLPerf closed, audited · B documented third-party method · C community / review · D vendor-published, not independently verified · E estimate. Power basis: measured with the result · spec vendor spec/envelope · power n/a not published.
Machine-readable: /find.json (see llms.txt); full tables are in “Show data”. Confidence grades: A audited MLPerf closed · B documented third-party · C community · D vendor-published · E estimate. Unknown values are shown as unknown, never estimated. Generated 2026-10-16 from public sources.