DeepSeek V4 Pro 1.6T — B300 vs MI325X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) on DeepSeek V4 Pro 1.6T. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
B300 hits 2522 tok/s/GPU for $0.26 per million tokens at 70 tok/s/user on DeepSeek V4 Pro 1.6T. No MI325X data at this operating point.
B300: 1133 tok/s/GPU, $0.57 per million tokens at 137 tok/s/user on DeepSeek V4 Pro 1.6T. MI325X is unmeasured here.
At 203 tok/s/user on DeepSeek V4 Pro 1.6T, B300 delivers 421 tok/s/GPU at $1.55 per million tokens; MI325X hasn't been benchmarked at this target. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:2521.7MI325X:— | B300:1132.8MI325X:— | B300:420.6MI325X:— |
| Cost ($/M tok) | B300:$0.262MI325X:— | B300:$0.572MI325X:— | B300:$1.549MI325X:— |
| tok/s/MW | B300:1327233MI325X:— | B300:596225MI325X:— | B300:221352MI325X:— |
| Concurrency | B300:~27MI325X:— | B300:~4MI325X:— | B300:~1MI325X:— |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.