Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on GLM 5/5.1. 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.
H200 posts 852 tok/s/chip for $0.40 per million tokens at 20 tok/s/user on GLM 5/5.1; MI325X posts 282 tok/s/chip for $1.08. H200 is 172% cheaper per token; H200 delivers 202% more tok/s/chip.
Throughput at 23 tok/s/user on GLM 5/5.1: H200 hits 852 tok/s/chip, MI325X hits 219. Per-million costs land at $0.40 and $1.40 respectively. H200 is 251% cheaper per token; H200 delivers 290% more tok/s/chip.
H200 / MI325X on GLM 5/5.1 at 26 tok/s/user: 838 / 159 tok/s/chip, $0.40 / $1.92 per million tokens. H200 is 375% cheaper per token; H200 delivers 426% more tok/s/chip. (Numbers reflect the default 8k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:851.9MI325X:282.1 | H200:851.9MI325X:218.6 | H200:838.1MI325X:159.2 |
| Cost ($/M tok) | H200:$0.398MI325X:$1.083 | H200:$0.398MI325X:$1.398 | H200:$0.404MI325X:$1.919 |
| tok/s/MW | H200:621824MI325X:166947 | H200:621824MI325X:129326 | H200:611771MI325X:94197 |
| Concurrency | H200:~32MI325X:~13 | H200:~32MI325X:~9 | H200:~31MI325X:~6 |
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
GLM5/5.1 744B 8K / 1K Total Tokens per $1 TCO vs. Interactivity