MiniMax M3 428B — B200 vs MI355X Performance per Dollar
Cost per million tokens of B200 (NVIDIA Blackwell) versus MI355X (AMD CDNA 4) on MiniMax M3 428B. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
Push MiniMax M3 428B to 95 tok/s/user and B200 lands at $0.33 per million tokens against MI355X's $0.28 — MI355X pulls ahead by 20%.
B200: $0.79 per million tokens. MI355X: $0.72. Both at 179 tok/s/user on MiniMax M3 428B, with MI355X 10% cheaper.
Toward the upper edge of the 11–348 tok/s/user interactivity band — at 264 tok/s/user — B200 runs $2.05 per million tokens on MiniMax M3 428B while MI355X runs $1.97. MI355X is the cheaper choice by 4%. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Chip pricing (owning hyperscaler): B200 $1.73/chip/hr · MI355X $1.50/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | B200:$0.331MI355X:$0.277 | B200:$0.787MI355X:$0.718 | B200:$2.052MI355X:$1.972 |
| Concurrency | B200:~46MI355X:~35 | B200:~7MI355X:~7 | B200:~2MI355X:~2 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.