32 × NVIDIA GB200

Tested system

Processor: NVIDIA Grace CPU (Arm Neoverse-V2)
Graphics: NVIDIA GB200
Accelerator count: 32
Nodes: 8

Recorded results

Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.

MLPerf Inference / deepseek-r1Offline · v6.1 · Tokens/s

32 × NVIDIA GB200

245,352^* Tokens/s

Accuracy target / result: exact_match: 81.8824065633546 TOKENS_PER_SAMPLE: 3798.4737921604374
Weight precision: fp4
Division: open
System type: datacenter
Availability: available
Cores per processor: 144
Processors per node: 2
Nodes: 8
Total accelerators: 32
Software stack: TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0
Operating system: Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k

Processing & platform

Processor: NVIDIA Grace CPU (Arm Neoverse-V2)
Graphics: NVIDIA GB200

Connectivity & software

Operating system: Ubuntu 24.04.4 LTS, Linux 6.14.0-37-generic-64k

System configuration

Reported host core count: 144
Processors per node: 2
Accelerator count: 32
Nodes: 8
Software stack: TensorRT 10.14, CUDA 13.1, TensorRT-LLM 1.3.0rc0
Submission division: open

* Provisional data