Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and H200 (NVIDIA Hopper) on MiniMax M2.5/M2.7. 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.
At 44 tok/s/user interactivity on MiniMax M2.5/M2.7, H100 delivers 1891 tok/s/chip at $0.17 per million tokens; H200 delivers 2766 tok/s/chip at $0.12. H200 is 40% cheaper per token; H200 delivers 46% more tok/s/chip at this point.
H100 posts 1269 tok/s/chip for $0.26 per million tokens at 66 tok/s/user on MiniMax M2.5/M2.7; H200 posts 2038 tok/s/chip for $0.17. H200 is 54% cheaper per token; H200 delivers 61% more tok/s/chip.
Throughput at 89 tok/s/user on MiniMax M2.5/M2.7: H100 hits 806 tok/s/chip, H200 hits 1367. Per-million costs land at $0.40 and $0.25 respectively. H200 is 63% cheaper per token; H200 delivers 70% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H100:1891.3H200:2766.1 | H100:1268.8H200:2037.5 | H100:805.9H200:1367.4 |
| Cost ($/M tok) | H100:$0.172H200:$0.123 | H100:$0.256H200:$0.166 | H100:$0.403H200:$0.248 |
| tok/s/MW | H100:1380510H200:2019085 | H100:926145H200:1487257 | H100:588262H200:998121 |
| Concurrency | H100:~39H200:~29 | H100:~18H200:~14 | H100:~9H200:~7 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
MiniMax M2.5/2.7 230B 8K / 1K Total Tokens per $1 TCO vs. Interactivity