MiniMax M3 428B — H100 vs H200
Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and H200 (NVIDIA Hopper) 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.
H100 / H200 on MiniMax M3 428B at 76 tok/s/user: 537 / 699 tok/s/GPU, $0.67 / $0.55 per million tokens. H200 is 22% cheaper per token; H200 delivers 30% more tok/s/GPU.
Around the middle of the 15–260 tok/s/user interactivity band, at 138 tok/s/user on MiniMax M3 428B: H100 runs 237 tok/s/GPU at $1.52/M tokens, H200 runs 360 at $1.06/M. H200 is 43% cheaper per token; H200 delivers 52% more tok/s/GPU.
Setting 199 tok/s/user as the target on MiniMax M3 428B, H100 produces 157 tok/s/GPU ($2.31 per million tokens) and H200 produces 223 ($1.73). H200 is 33% cheaper per token; H200 delivers 42% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H100:536.7H200:699.4 | H100:236.6H200:359.6 | H100:156.6H200:223.0 |
| Cost ($/M tok) | H100:$0.673H200:$0.551 | H100:$1.518H200:$1.062 | H100:$2.307H200:$1.731 |
| tok/s/MW | H100:310239H200:404301 | H100:136749H200:207872 | H100:90500H200:128925 |
| Concurrency | H100:~31H200:~20 | H100:~8H200:~5 | H100:~4H200:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.