MiniMax M3.1 Flash 428B · Chip comparison
MiniMax M3.1 Flash 428B — H100 vs Vera Rubin NVL72
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and Vera Rubin NVL72 (NVIDIA Vera Rubin) on MiniMax M3.1 Flash 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 313 tok/s/user as the target on MiniMax M3.1 Flash 428B, H100 produces 2,352 tok/s/chip ($0.14 per million tokens) and Vera Rubin NVL72 produces 254,869 ($0.004). Vera Rubin NVL72 is 97% cheaper per token; Vera Rubin NVL72 delivers 108× the tok/s/chip.
At 610 tok/s/user interactivity on MiniMax M3.1 Flash 428B, H100 delivers 2,352 tok/s/chip at $0.14 per million tokens; Vera Rubin NVL72 delivers 61,404 tok/s/chip at $0.016. Vera Rubin NVL72 is 88% cheaper per token; Vera Rubin NVL72 delivers 26× the tok/s/chip at this point.
H100 posts 2,352 tok/s/chip for $0.14 per million tokens at 908 tok/s/user on MiniMax M3.1 Flash 428B; Vera Rubin NVL72 posts 32,066 tok/s/chip for $0.031. Vera Rubin NVL72 is 77% cheaper per token; Vera Rubin NVL72 delivers 14× the tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H1002,352Vera Rubin NVL72254,869 | H1002,352Vera Rubin NVL7261,404 | H1002,352Vera Rubin NVL7232,066 |
| Cost ($/M tok) | H100$0.14Vera Rubin NVL72$0.004 | H100$0.14Vera Rubin NVL72$0.016 | H100$0.14Vera Rubin NVL72$0.031 |
| tok/s/MW | H1001,716,747Vera Rubin NVL7277,232,975 | H1001,716,747Vera Rubin NVL7218,607,346 | H1001,716,747Vera Rubin NVL729,716,938 |
| Concurrency | H100~1Vera Rubin NVL72~640 | H100~1Vera Rubin NVL72~61 | H100~1Vera Rubin NVL72~11 |