MiniMax M3 428B · Chip comparison
MiniMax M3 428B — GB200 NVL72 vs MI355X
Head-to-head AI inference benchmark comparison of GB200 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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Setting 168 tok/s/user as the target on MiniMax M3 428B, GB200 NVL72 produces 33,531 tok/s/chip ($0.015 per million tokens) and MI355X produces 38,982 ($0.011). MI355X is 31% cheaper per token; MI355X delivers 16% more tok/s/chip.
At 253 tok/s/user interactivity on MiniMax M3 428B, GB200 NVL72 delivers 16,343 tok/s/chip at $0.032 per million tokens; MI355X delivers 19,743 tok/s/chip at $0.021. MI355X is 33% cheaper per token; MI355X delivers 21% more tok/s/chip at this point.
GB200 NVL72 posts 8,176 tok/s/chip for $0.063 per million tokens at 338 tok/s/user on MiniMax M3 428B; MI355X posts 9,858 tok/s/chip for $0.042. MI355X is 33% cheaper per token; MI355X delivers 21% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL7233,531MI355X38,982 | GB200 NVL7216,343MI355X19,743 | GB200 NVL728,176MI355X9,858 |
| Cost ($/M tok) | GB200 NVL72$0.015MI355X$0.011 | GB200 NVL72$0.032MI355X$0.021 | GB200 NVL72$0.063MI355X$0.042 |
| tok/s/MW | GB200 NVL7217,930,819MI355X18,651,786 | GB200 NVL728,739,563MI355X9,446,261 | GB200 NVL724,372,272MI355X4,716,870 |
| Concurrency | GB200 NVL72~19MI355X~21 | GB200 NVL72~9MI355X~12 | GB200 NVL72~2MI355X~5 |