MiniMax M3 428B — GB300 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on MiniMax M3 428B. 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.
Near the low end of the 22–307 tok/s/user interactivity band, at 93 tok/s/user on MiniMax M3 428B: GB300 NVL72 runs 12281 tok/s/chip at $0.05/M tokens, MI355X runs 3912 at $0.11/M. GB300 NVL72 is 102% cheaper per token; GB300 NVL72 delivers 214% more tok/s/chip.
Setting 164 tok/s/user as the target on MiniMax M3 428B, GB300 NVL72 produces 2708 tok/s/chip ($0.26 per million tokens) and MI355X produces 2112 ($0.20). MI355X is 34% cheaper per token; GB300 NVL72 delivers 28% more tok/s/chip.
At 236 tok/s/user interactivity on MiniMax M3 428B, GB300 NVL72 delivers 1756 tok/s/chip at $0.36 per million tokens; MI355X delivers 1238 tok/s/chip at $0.33. MI355X is 11% cheaper per token; GB300 NVL72 delivers 42% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:12281.5MI355X:3911.5 | GB300 NVL72:2708.2MI355X:2112.4 | GB300 NVL72:1756.4MI355X:1237.9 |
| Cost ($/M tok) | GB300 NVL72:$0.052MI355X:$0.106 | GB300 NVL72:$0.264MI355X:$0.197 | GB300 NVL72:$0.363MI355X:$0.327 |
| tok/s/MW | GB300 NVL72:5793144MI355X:1871534 | GB300 NVL72:1277447MI355X:1010713 | GB300 NVL72:828513MI355X:592275 |
| Concurrency | GB300 NVL72:~672MI355X:~22 | GB300 NVL72:~20MI355X:~7 | GB300 NVL72:~8MI355X:~3 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.