Llama 3.1 8B 3.947 min 6.1-0133 closed Available on-premise NVIDIAHecate (8x NVIDIA VR NVL72) 512 128 NVIDIA Rubin GPU 4 Olympus 256 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 4.459 min 6.0-0015 closed Available Cloud NVIDIATyche-hsg (16x NVIDIA GB200 NVL72) 1,024 256 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 512 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 4.636 min 6.0-0101 closed Available on-premise NVIDIATheia (8x NVIDIA GB300 NVL72) 512 128 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 256 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 5.219 min 5.1-0060 closed Available on-premise NVIDIATheia (8x NVIDIA GB300 NVL72) 512 128 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 256 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 5.432 min 5.1-0072 closed Available on-premise NVIDIATyche (8x NVIDIA GB200 NVL72) 512 128 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 256 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 7.162 min 6.1-0073 closed Available on-premise CoreWeaveCoreWeave_VR200_18x4 72 18 NVIDIA Rubin GPU 4 Olympus 36 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 7.380 min 6.1-0123 closed Available on-premise NVIDIAHecate (1x NVIDIA VR NVL72) 72 18 NVIDIA Rubin GPU 4 Olympus 36 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 7.683 min 6.1-0140 closed Available on-premise SupermicroSRS-VR-NVL72 (18x ARS-122GL-R2-ALC) 72 18 NVIDIA Rubin GPU 4 Olympus 36 NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 8.784 min 6.1-0148 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 256 32 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 64 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 10.024 min 5.1-0087 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 256 32 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 64 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 11.541 min 6.1-0138 closed Available on-premise Quanta_Cloud_TechnologyD75U-1U_ngpu72 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 11.569 min 5.1-0085 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 128 16 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 32 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 11.586 min 6.0-0009 closed Available Cloud LambdaLambda_GB300_n18 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 PyTorch NVIDIA Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 11.599 min 6.1-0146 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 128 16 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 32 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 11.753 min 6.0-0096 closed Available on-premise NVIDIATheia (1x NVIDIA GB300 NVL72) 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 11.821 min 6.0-0077 closed Available on-premise HPENVIDIA GB300 NVL72 by HPE 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA Release 26.04 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 11.844 min 6.0-0063 closed Available on-premise GigaComputingDLB2-CB3 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 11.860 min 6.0-0110 closed Available on-premise SupermicroSRS-GB300-NVL72-M1 (18x ARS-121GL-NB3) 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NeMo Framework NVIDIA Release 26.04; NeMo Framework NVIDIA Release 26.03 Implementation v6.0 v6.0 Llama 3.1 8B 11.872 min 6.0-0026 closed Available Cloud NebiusNebius GB300 n18 (18x NVIDIA GB300 NVL4) 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 Implementation v6.0 v6.0 Llama 3.1 8B 11.884 min 6.1-0098 closed Available on-premise HPEGB300 NVL72 for HPE (1x NVIDIA GB300 NVL72) 64 16 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 32 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 12.068 min 6.0-0030 closed Available Cloud OracleBM.GPU.GB300.4 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 12.159 min 6.1-0137 closed Available on-premise Quanta_Cloud_TechnologyD75U-1U_ngpu64 64 16 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 32 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 12.447 min 6.0-0104 closed Available on-premise Quanta_Cloud_TechnologyD75U-1U_ngpu64 64 16 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 32 Implementation v6.0 v6.0 Llama 3.1 8B 12.686 min 6.0-0045 closed Available on-premise Dell16xXE9712x4GB300 64 16 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 32 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 12.686 min 6.1-0075 closed Available on-premise Dell16xXE9712x4GB300 64 16 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 32 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 12.976 min 5.1-0057 closed Available on-premise NVIDIATheia (1x NVIDIA GB300 NVL72) 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 13.701 min 6.1-0070 closed Available on-premise CoreWeaveCoreWeave_GB200_18x4 72 18 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 36 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 14.074 min 6.1-0003 closed Available Cloud Crusoe16xMI355X-288GB_2xEPYC-9575F 128 16 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 32 flux1: PyTorch (ROCm 7.0.2) + TransformerEngine (fp8) MLPerf flux1 reference harness (torchtitan/NeMo); llama: AMD Primus (Megatron-LM pretrain / Megatron-Bridge post-train) from rocm/amd-mlperf:llama31_8b_training_6.0 and rocm/amd-mlperf:llama2_70b_training_6.0 (PyTorch 2.10 ROCm TransformerEngine ROCm AITER) Implementation v6.1 v6.1 Llama 3.1 8B 14.123 min 6.1-0069 closed Available on-premise CiscoZeus_8xB300_SXM_288GB_multi_planar 64 8 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 16 PyTorch NVIDIA Release 26.04 (Llama and GPT-OSS); 26.09 (DeepSeek and DLRMv4); 26.08 (Qwen) NVIDIA NeMo / Megatron TensorRT-LLM (Qwen generation) TorchRec (DLRMv4) Implementation v6.1 v6.1 Llama 3.1 8B 14.255 min 5.1-0001 closed Available Cloud LambdaLambda_GB300_n18 72 18 NVIDIA Blackwell Ultra GPU (GB300) 4 Intel(R) Xeon(R) Platinum 8570 36 PyTorch NVIDIA Release 25.04 Implementation v5.1 v5.1 Llama 3.1 8B 14.416 min 6.0-0025 closed Available Cloud NebiusNebius B300 n8 (64x B300-SXM-270GB) 64 8 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 16 NVIDIA NeMo Framework Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 15.020 min 5.1-0063 closed Available on-premise NVIDIATyche (1x NVIDIA GB200 NVL72) 72 18 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 36 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 16.127 min 6.1-0150 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 16 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 16.536 min 6.0-0002 closed Available Cloud CoreWeaveCoreWeave_B200_8x8 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8562Y+ 16 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 18.100 min 5.1-0090 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 16 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 18.256 min 5.1-0013 closed Available Cloud OracleBM.GPU.B200.8 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 INTEL(R) XEON(R) PLATINUM 8592+ 16 MXNet NVIDIA Release 24.04 or PyTorch NVIDIA Release 24.04 Implementation v5.1 v5.1 Llama 3.1 8B 18.305 min 6.0-0034 closed Available Cloud SCITIXSCITIX.B200x8.N8 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) 6767P 16 PyTorch NVIDIA Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 18.859 min 5.1-0028 closed Available on-premise VerdaDatacrunch-Instant-Cluster-8xB200 64 8 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9575F 16 NeMo 2.1.0 Megatron-LM PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 20.200 min 6.0-0060 closed Available on-premise Dell8xXE9712x4GB300 32 8 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 16 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 20.200 min 6.1-0090 closed Available on-premise Dell8xXE9712x4GB300 32 8 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 16 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 21.433 min 6.0-0076 closed Available on-premise HPENVIDIA GB300 NVL72 by HPE 32 8 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 16 NVIDIA Release 26.04 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 21.975 min 6.0-0042 closed Available on-premise CiscoZeus_16xH100_SXM_80GB 128 16 NVIDIA H100-SXM5-80GB 8 Intel(R) Xeon(R) Platinum 8592+ 32 PyTorch NVIDIA NGC container Implementation v6.0 v6.0 Llama 3.1 8B 24.748 min 6.1-0149 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 32 4 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 8 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 27.828 min 5.1-0007 closed Available Cloud NebiusNebius B200 n4 (32x B200-SXM-180GB) 32 4 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8580 8 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 32.896 min 5.1-0026 closed Available on-premise Ciscohpf_n16 128 16 NVIDIA H100-SXM5-80GB 8 Intel(R) Xeon(R) Platinum 8592+ 32 PyTorch NVIDIA Release 23.09 Implementation v5.1 v5.1 Llama 3.1 8B 33.391 min 6.0-0075 closed Available on-premise HPENVIDIA GB300 NVL72 by HPE 16 4 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 8 NVIDIA Release 26.04 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 33.808 min 6.0-0044 closed Available on-premise CiscoZeus_8xH100_SXM_80GB 64 8 NVIDIA H100-SXM5-80GB 8 Intel(R) Xeon(R) Platinum 8592+ 16 PyTorch NVIDIA NGC container Implementation v6.0 v6.0 Llama 3.1 8B 35.579 min 6.0-0059 closed Available on-premise Dell4xXE9712x4GB300 16 4 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 8 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 35.579 min 6.1-0089 closed Available on-premise Dell4xXE9712x4GB300 16 4 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 8 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 35.682 min 5.1-0088 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 32 4 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 8 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 36.242 min 6.1-0125 closed Available on-premise NVIDIAHecate (1x NVIDIA VR NVL72) 8 2 NVIDIA Rubin GPU 4 Olympus 4 NVIDIA NeMo 26.09 Plus Implementation v6.1 v6.1 Llama 3.1 8B 37.753 min 5.1-0043 closed Available on-premise HPEHPE Cray XD670 (8x H200-SXM-141GB) 64 8 NVIDIA H200-SXM5-141GB 8 Intel(R) Xeon(R) Platinum 8562Y+ 16 NVIDIA Release 25.08 (NeMo PyTorch DGL HugeCTR) Implementation v5.1 v5.1 Llama 3.1 8B 39.014 min 6.0-0072 closed Available on-premise HPENVIDIA GB200 NVL72 by HPE 32 8 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 16 NVIDIA Release 25.09 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 40.964 min 6.1-0100 closed Available on-premise HPEHPE Compute XD690 (8x B300-SXM6-288GB) 16 2 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 4 NVIDIA Release 26.04 (NeMo PyTorch DGL HugeCTR) Implementation v6.1 v6.1 Llama 3.1 8B 43.226 min 6.1-0005 closed Available Cloud Crusoe2xMI355X-288GB_2xEPYC-9575F 16 2 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 4 gpt_oss_20b: AMD Primus (Megatron-LM pretrain) from rocm/amd-mlperf:llama31_8b_training_6.0 (PyTorch 2.10 ROCm TransformerEngine ROCm AITER); dlrmv4: PyTorch + TorchRec/FBGEMM-GPU (generative-recommenders HSTU) AMD MLPerf Training v6.1 dlrmv4 recipe; llama2_70b_lora: AMD Primus (Megatron-Bridge post-train) from rocm/amd-mlperf:llama2_70b_training_6.0 (PyTorch 2.10 ROCm TransformerEngine ROCm AITER); llama31_8b: AMD Primus (Megatron-LM pretrain) AMD MLPerf Training v6.1 llama31_8b recipe Implementation v6.1 v6.1 Llama 3.1 8B 43.317 min 6.0-0058 closed Available on-premise Dell2xXE9780x8B300-SXM-270GB 16 2 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6747P 4 PyTorch NVIDIA Release 26.03 Implementation v6.0 v6.0 Llama 3.1 8B 43.317 min 6.1-0088 closed Available on-premise Dell2xXE9780x8B300-SXM-270GB 16 2 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 4 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 44.104 min 6.1-0147 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 16 2 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 4 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 45.445 min 5.1-0027 closed Available on-premise Ciscohpf_n8 64 8 NVIDIA H100-SXM5-80GB 8 Intel(R) Xeon(R) Platinum 8592+ 16 PyTorch NVIDIA Release 23.09 Implementation v5.1 v5.1 Llama 3.1 8B 45.858 min 6.0-0040 closed Available on-premise CiscoCisco UCS C880A-8xB300-SXM-288G 16 2 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 4 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 49.047 min 6.0-0071 closed Available on-premise HPENVIDIA GB200 NVL72 by HPE 16 4 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 8 NVIDIA Release 25.09 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 51.827 min 5.1-0006 closed Available Cloud NebiusNebius B200 n2 (16x B200-SXM-180GB) 16 2 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8580 4 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 52.263 min 6.0-0069 closed Available on-premise HPEHPE ProLiant Compute XD685 16 2 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9535 4 NVIDIA Framework Release 25.09 (NeMo PyTorch DGL HugeCTR) Implementation v6.0 v6.0 Llama 3.1 8B 52.402 min 6.1-0094 closed Available on-premise FirmusFirmus Supermicro GB300 NVL72 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 55.050 min 6.0-0109 closed Available on-premise SupermicroAS-8126GS-NB3RT 16 2 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 AMD EPYC 9575F 4 NeMo Framework NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 56.074 min 6.1-0152 closed Available on-premise tinycorptinybox 2x8xMI350X 16 2 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 4 tinygrad Implementation v6.1 v6.1 Llama 3.1 8B 58.629 min 6.0-0043 closed Available on-premise CiscoZeus_2xMI350X 16 2 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 64-Core Processor 4 PyTorch + ROCm (Primus / NeMo for LLM workloads) Implementation v6.0 v6.0 Llama 3.1 8B 59.561 min 5.1-0086 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 16 2 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 4 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 60.811 min 5.1-0038 closed Available on-premise Dell2xXE9685Lx8B200-SXM-180GB 16 2 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9965 4 NVIDIA NGC Pytorch 25.04 HugeCTR Nemo 25.09 DGL 25.03 Implementation v5.1 v5.1 Llama 3.1 8B 62.053 min 6.1-0054 closed Available Cloud OracleBM.GPU.GB300.4 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA PyTorch/NeMo Releases 26.09 and 26.10 Implementation v6.1 v6.1 Llama 3.1 8B 62.643 min 6.1-0139 closed Available on-premise Quanta_Cloud_TechnologyD75U-1U_ngpu8 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA NeMo Framework Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 63.516 min 6.0-0078 closed Available on-premise HPENVIDIA GB300 NVL72 by HPE 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA Release 26.04 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 63.593 min 6.0-0105 closed Available on-premise Quanta_Cloud_TechnologyD75U-1U_ngpu8 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 Implementation v6.0 v6.0 Llama 3.1 8B 63.678 min 6.0-0057 closed Available on-premise Dell2xXE9712x4GB300 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 63.678 min 6.1-0087 closed Available on-premise Dell2xXE9712x4GB300 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 64.346 min 6.0-0027 closed Available Cloud NebiusNebius GB300 n2 (2x NVIDIA GB300 NVL4) 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 Implementation v6.0 v6.0 Llama 3.1 8B 65.258 min 6.0-0031 closed Available Cloud OracleBM.GPU.GB300.4 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA NeMo Framework Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 66.042 min 6.0-0098 closed Available on-premise NVIDIATheia (1x NVIDIA GB300 NVL72) 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 67.373 min 5.1-0058 closed Available on-premise NVIDIATheia (1x NVIDIA GB300 NVL72) 8 2 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 4 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 69.679 min 6.1-0096 closed Available on-premise GigaComputingG4L4-AD3-LAX7 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6979P 2 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 71.320 min 6.1-0068 closed Available on-premise ASUSTeKXA_NB3I-E12 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 NVIDIA Release pytorch 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 71.496 min 6.1-0151 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B300 (8x B300-SXM-270GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 2 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 71.600 min 6.1-0084 closed Available on-premise Dell1xXE9785Lx8B300-SXM-270GB 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 AMD EPYC 9965 2 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 72.012 min 6.0-0023 closed Available Cloud NebiusNebius B300 n1 (8x B300-SXM-270GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 2 NVIDIA NeMo Framework Release 25.09 Implementation v6.0 v6.0 Llama 3.1 8B 72.018 min 6.1-0103 closed Available on-premise InventecInventec P5800G7(LC) 1x (8x B300-SXM-270GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6787P 2 NVIDIA Megatron-Bridge 26.08 Implementation v6.1 v6.1 Llama 3.1 8B 72.942 min 6.1-0062 closed Available on-premise AMDSupermicro AS-4126GS-NMR-LCC 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 v26.6_primus Implementation v6.1 v6.1 Llama 3.1 8B 73.486 min 6.0-0038 closed Available on-premise ASUSTeKXA_NB3I-E12 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 NVIDIA Release pytorch 26.0426.03 Implementation v6.0 v6.0 Llama 3.1 8B 73.691 min 6.1-0008 closed Available Cloud CrusoeCrusoe Cloud B300 n1 (8x B300-SXM-288GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 2 NVIDIA NeMo Framework (Megatron-Bridge) built on NGC PyTorch 26.04 container (nvcr.io/nvidia/pytorch:26.04-py3) Implementation v6.1 v6.1 Llama 3.1 8B 74.081 min 6.0-0066 closed Available on-premise GigaComputingG894-SD3-AAX7 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 PyTorch Implementation v6.0 v6.0 Llama 3.1 8B 74.255 min 6.0-0053 closed Available on-premise Dell1xXE9780x8B300-SXM-270GB 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6787P 2 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 74.751 min 6.0-0039 closed Available on-premise CiscoCisco UCS C880A-8xB300-SXM-288G 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 74.787 min 6.0-0106 closed Available on-premise SupermicroAS-4126GS-NB3RT-LCC 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 AMD EPYC 9575F 2 NeMo Framework NVIDIA Release 26.04; Merlin HugeCTR NVIDIA Release 26.03 Implementation v6.0 v6.0 Llama 3.1 8B 74.958 min 6.0-0067 closed Available on-premise HPEHPE Compute XD690 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 2 NVIDIA Release 26.04 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 75.132 min 6.1-0086 closed Available on-premise Dell1xXE9785x8MI355X 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9655 2 Pytorch 25.04 Implementation v6.1 v6.1 Llama 3.1 8B 75.808 min 6.0-0103 closed Available on-premise Quanta_Cloud_TechnologyD75H-10U 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6740P 2 Implementation v6.0 v6.0 Llama 3.1 8B 75.841 min 5.1-0008 closed Available Cloud NebiusNebius B300 n1 (8x B300-SXM-270GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6776P 2 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 76.165 min 6.1-0141 closed Available on-premise SupermicroSupermicro AS -A126GS-TNMR (8x AMD Instinct MI355X) 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 AMD Primus (PyTorch 2.12.0 ROCm 7.15.0) Implementation v6.1 v6.1 Llama 3.1 8B 76.407 min 6.1-0102 closed Available on-premise HPEHPE ProLiant Compute XD685 (8x MI355X) 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9555 2 AMD Primus: Megatron-LM + Primus-Turbo (Llama 3.1 8B GPT-OSS 20B) Megatron-Bridge + Primus-Turbo (Llama 2 70B LoRA) all AMD MLPerf 6.1 partner code 253c561; AMD MLPerf v6.0 FLUX.1 implementation (FLUX.1); PyTorch + TorchRec + FBGEMM generative-recommenders HSTU (DLRMv4: AMD MLPerf 6.1 partner code 253c561) Implementation v6.1 v6.1 Llama 3.1 8B 76.768 min 5.1-0050 closed Available on-premise LenovoThinkSystem SR680a V4 with 8x B300-SXM 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6747P 2 PyTorch NVIDIA Release 23.09 Implementation v5.1 v5.1 Llama 3.1 8B 77.328 min 6.0-0079 closed Available on-premise InventecP9000IG7 (8x B300-SXM-270GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 NVIDIA Megatron-Bridge 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 77.393 min 6.1-0101 closed Available on-premise HPEHPE ProLiant Compute XD685 (8x B300-SXM6-288GB) 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 AMD EPYC 9555 2 NVIDIA Framework Release 26.04 (NeMo PyTorch DGL HugeCTR) Implementation v6.1 v6.1 Llama 3.1 8B 78.339 min 6.1-0004 closed Available Cloud Crusoe1xMI355X-288GB_2xEPYC-9575F 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 llama: AMD Primus (Megatron-Bridge post-train) from rocm/amd-mlperf:llama2_70b_training_6.0 (PyTorch 2.10 ROCm TransformerEngine ROCm AITER); dlrmv4: PyTorch + TorchRec/FBGEMM-GPU (generative-recommenders HSTU) AMD MLPerf Training v6.1 dlrmv4 recipe; llama31_8b: AMD Primus (Megatron-LM pretrain) AMD MLPerf Training v6.1 llama31_8b recipe Implementation v6.1 v6.1 Llama 3.1 8B 79.325 min 5.1-0067 closed Available on-premise NVIDIATyche (1x NVIDIA GB200 NVL72) 8 2 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 4 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 79.692 min 6.0-0074 closed Available on-premise HPENVIDIA GB200 NVL72 by HPE 8 2 NVIDIA Blackwell GPU (GB200) 4 Neoverse-V2 4 NVIDIA Release 25.09 (NeMo PyTorch) Implementation v6.0 v6.0 Llama 3.1 8B 80.907 min 6.1-0083 closed Available on-premise Dell1xXE9785Lx8B200-SXM-180GB 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9655 2 PyTorch NVIDIA Release 26.09 Implementation v6.1 v6.1 Llama 3.1 8B 81.894 min 6.1-0145 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 2 PyTorch NVIDIA Release 26.09 (Flux.1: 26.08) Implementation v6.1 v6.1 Llama 3.1 8B 82.213 min 6.0-0052 closed Available on-premise Dell1xXE9780x8B200-SXM-180GB 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) 6767P 2 DGL NVIDIA Release 25.03 Implementation v6.0 v6.0 Llama 3.1 8B 83.680 min 6.0-0107 closed Available on-premise SupermicroAS-4126GS-NBR-LCC 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9965 2 NeMo Framework NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 84.379 min 5.1-0081 closed Available on-premise SupermicroSYS-422GS-NBRT-LCC 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) 6787P 2 PyTorch NVIDIA Release 25.04; NeMo Framework NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 84.834 min 5.1-0011 closed Available Cloud OracleBM.GPU.B200.8 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 INTEL(R) XEON(R) PLATINUM 8592+ 2 MXNet NVIDIA Release 24.04 or PyTorch NVIDIA Release 24.04 Implementation v5.1 v5.1 Llama 3.1 8B 85.027 min 5.1-0030 closed Available on-premise Dell1xXE9680Lx8B200-SXM-180GB 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8562Y+ 2 NVIDIA NGC Pytorch 25.04 HugeCTR Nemo 25.09 DGL 25.03 Implementation v5.1 v5.1 Llama 3.1 8B 85.204 min 5.1-0051 closed Available on-premise LenovoThinkSystem SR780a V3 with 8x B200-SXM-180GB 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8562Y+ 2 PyTorch NVIDIA Release 23.09 Implementation v5.1 v5.1 Llama 3.1 8B 85.263 min 6.0-0008 closed Available Cloud LambdaLambda-1-Click-Cluster_B200_n1 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 2 PyTorch NVIDIA Release 25.04 Implementation v6.0 v6.0 Llama 3.1 8B 85.369 min 5.1-0005 closed Available Cloud NebiusNebius B200 n1 (8x B200-SXM-180GB) 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8580 2 NVIDIA NeMo Framework Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 85.371 min 6.0-0068 closed Available on-premise HPEHPE ProLiant Compute XD685 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 AMD EPYC 9535 2 NVIDIA Framework Release 25.09 (NeMo PyTorch DGL HugeCTR) Implementation v6.0 v6.0 Llama 3.1 8B 86.845 min 6.0-0036 closed Available on-premise AMDsmci355-ccs-aus-n12-05 (AS -4126GS-NMR-LCC) 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 AMD Primus Framework 0.2.0 Implementation v6.0 v6.0 Llama 3.1 8B 90.037 min 6.0-0108 closed Available on-premise SupermicroAS-4126GS-NMR-LCC 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 PyTorch 2.10.0+rocm7.2.2.git40d237bf ROCm 7.2.2 Implementation v6.0 v6.0 Llama 3.1 8B 90.057 min 6.0-0116 closed Available on-premise vultrSMCI355X (AS -4126GS-NMR-LCC) 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 PyTorch 2.10.0a0+git449b176 ROCm 7.2.0 Implementation v6.0 v6.0 Llama 3.1 8B 91.145 min 6.0-0033 closed Available Cloud OracleBM.GPU.MI355X.8 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 ROCm 7.0 Implementation v6.0 v6.0 Llama 3.1 8B 91.251 min 6.0-0064 closed Available on-premise GigaComputingG893-ZX1-AAX4 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 PyTorch Implementation v6.0 v6.0 Llama 3.1 8B 91.413 min 6.0-0055 closed Available on-premise Dell1xXE9785Lx8MI355X 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9655 2 Pytorch 25.04 Implementation v6.0 v6.0 Llama 3.1 8B 91.939 min 6.0-0070 closed Available on-premise HPEHPE ProLiant Compute XD685 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9575F 2 Implementation v6.0 v6.0 Llama 3.1 8B 92.075 min 6.1-0067 closed Available on-premise AMDSupermicro AS-8126GS-TNMR 8 1 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 2 v26.6_primus Implementation v6.1 v6.1 Llama 3.1 8B 92.829 min 6.0-0090 closed Available on-premise MiTACG4826 8 1 AMD Instinct MI355X 288GB HBM3e 8 AMD EPYC 9755 2 Pytorch 2.9.1 ROCm 7.2.2 Implementation v6.0 v6.0 Llama 3.1 8B 93.687 min 5.1-0084 closed Available on-premise University-of-FloridaHiPerGator NVIDIA DGX B200 (8x B200-SXM-180GB) 8 1 NVIDIA Blackwell GPU (B200-SXM-180GB) 8 Intel(R) Xeon(R) Platinum 8570 2 PyTorch NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 94.817 min 5.1-0077 closed Available on-premise SupermicroAS-8126GS-NB3RT 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 AMD EPYC 9965 2 PyTorch NVIDIA Release 25.04; NeMo Framework NVIDIA Release 25.09 Implementation v5.1 v5.1 Llama 3.1 8B 95.383 min 5.1-0035 closed Available on-premise Dell1xXE9780x8B300-SXM-270GB 8 1 NVIDIA Blackwell Ultra GPU (B300-SXM-270GB) 8 Intel(R) Xeon(R) 6767P 2 NVIDIA NGC Pytorch 25.04 HugeCTR Nemo 25.09 DGL 25.03 Implementation v5.1 v5.1 Llama 3.1 8B 99.709 min 5.1-0018 closed Available on-premise AMDSMCI355X (AS -4126GS-NMR-LCC) 8 1 AMD Instinct MI355X 8 AMD EPYC 9575F 2 PyTorch 2.8.0.dev20251001+rocm7.0.0.git64359f59 ROCm 7.0.0 Implementation v5.1 v5.1 Llama 3.1 8B 101.278 min 6.1-0153 closed Available on-premise tinycorptinybox 8xMI350X 8 1 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 2 tinygrad Implementation v6.1 v6.1 Llama 3.1 8B 103.559 min 5.1-0076 closed Available on-premise SupermicroAS-4126GS-NMR-LCC 8 1 AMD Instinct MI355X 8 AMD EPYC 9575F 2 PyTorch ROCm 7.0.1 Implementation v5.1 v5.1 Llama 3.1 8B 105.667 min 5.1-0042 closed Available on-premise GigaComputingGIGABYTE G893-ZX1-AAX4 8 1 AMD Instinct MI355X 8 AMD EPYC 9575F 2 PyTorch Implementation v5.1 v5.1 Llama 3.1 8B 107.462 min 5.1-0036 closed Available on-premise Dell1xXE9785Lx8MI355X 8 1 AMD Instinct MI355X 8 AMD EPYC 9655 2 Pytorch 25.04 Implementation v5.1 v5.1 Llama 3.1 8B 108.965 min 6.0-0092 closed Available on-premise MiTACG8825Z5 8 1 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 2 Flux1: PyTorch 2.10.0+rocm7.2.2.git40d237bf Llama3.1-8B:Pytorch 2.10.0a0+git449b176 ROCM 7.2.0 Implementation v6.0 v6.0 Llama 3.1 8B 109.756 min 6.0-0035 closed Available on-premise AMDsmci350-rck-g03-b18-40 (AS -8126GS-TNMR) 8 1 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 2 AMD Primus Framework 0.2.0 Implementation v6.0 v6.0 Llama 3.1 8B 122.199 min 6.0-0041 closed Available on-premise CiscoCisco UCS C885A-8xH200-SXM-141G 16 2 NVIDIA H200-SXM5-141GB 8 AMD EPYC 9575F 4 NVIDIA NeMo Framework Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 122.929 min 5.1-0017 closed Available on-premise AMDSMCI350X (AS -8126GS-TNMR) 8 1 AMD Instinct MI350X 8 AMD EPYC 9575F 2 PyTorch 2.8.0.dev20251001+rocm7.0.0.git64359f59 ROCm 7.0.0 Implementation v5.1 v5.1 Llama 3.1 8B 123.732 min 6.0-0048 closed Available on-premise Dell1xXE9712x4GB300 4 1 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 2 PyTorch NVIDIA Release 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 123.732 min 6.1-0078 closed Available on-premise Dell1xXE9712x4GB300 4 1 NVIDIA Blackwell Ultra GPU (GB300) 4 Neoverse-V2 2 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 127.625 min 5.1-0025 closed Available on-premise CiscoCisco UCS C885A_M8-8xMI350X-OAM-288GB 8 1 AMD Instinct MI350X 8 AMD EPYC 9575F 2 PyTorch 2.8.0.dev20251001+rocm7.0.0.git64359f59 ROCm 7.0.0 Implementation v5.1 v5.1 Llama 3.1 8B 187.479 min 6.0-0115 closed Available on-premise tinycorptinybox 8xMI350X 8 1 AMD Instinct MI350X 288GB HBM3e 8 AMD EPYC 9575F 2 tinygrad branch mlperf_training_v6.0 Implementation v6.0 v6.0 Llama 3.1 8B 238.073 min 6.0-0093 closed Available on-premise MiTACMI325X_1node 8 1 AMD Instinct MI325X 256GB HBM3e 8 AMD EPYC 9575F 2 NeMo (flux1) / Primus + Megatron-LM (llama31_8b) on PyTorch ROCm Implementation v6.0 v6.0 Llama 3.1 8B 257.413 min 6.0-0046 closed Available on-premise Dell1xXE7740x8H200-NVL-141GB 8 1 NVIDIA H200-NVL-141GB 8 Intel(R) Xeon(R) 6767P 2 NVIDIA NGC Pytorch 26.04 Implementation v6.0 v6.0 Llama 3.1 8B 259.950 min 6.1-0077 closed Available on-premise Dell1xXE7740x8H200-NVL-141GB 8 1 NVIDIA H200-NVL-141GB 8 Intel(R) Xeon(R) 6767P 2 NVIDIA NGC Pytorch 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 274.708 min 6.1-0095 closed Available on-premise FujitsuPRIMERGY_CDI_H200_NVL_141GBx8 8 1 NVIDIA H200-NVL-141GB 8 Intel(R) Xeon(R) Platinum 8592+ 2 PyTorch NVIDIA Release 26.04 Implementation v6.1 v6.1 Llama 3.1 8B 375.636 min 6.0-0114 closed Available on-premise TTAKlimax-4082 8 1 NVIDIA RTX PRO 6000 Blackwell Server Edition 8 AMD EPYC 9654 2 NeMo 2.6.0rc0 + Megatron-LM core_r0.11.0 Implementation v6.0 v6.0 Llama 3.1 8B 1,187.380 min 6.1-0144 closed Available on-premise TTATTA 4x H100 NVL with ExaStor-MT parallel filesystem 4 1 H100 NVL Tensor Core GPU (PCIe) 4 Intel(R) Xeon(R) Platinum 8558 2 PyTorch NVIDIA Release 25.09 (NeMo) Implementation v6.1 v6.1