Head-to-head AI inference benchmark comparison of H100 (NVIDIA Hopper) and MI325X (AMD CDNA 3) on gpt-oss 120B. Latency, throughput, and cost across LLM workloads. Use the chart controls below to switch sequences, precisions, and metrics β same interactions as the main inference chart.
At 54 tok/s/user interactivity on gpt-oss 120B, H100 delivers 8610 tok/s/chip at $0.04 per million tokens; MI325X delivers 4542 tok/s/chip at $0.07. H100 is 78% cheaper per token; H100 delivers 90% more tok/s/chip at this point.
H100 posts 7306 tok/s/chip for $0.04 per million tokens at 72 tok/s/user on gpt-oss 120B; MI325X posts 3543 tok/s/chip for $0.09. H100 is 94% cheaper per token; H100 delivers 106% more tok/s/chip.
Throughput at 89 tok/s/user on gpt-oss 120B: H100 hits 6413 tok/s/chip, MI325X hits 1440. Per-million costs land at $0.05 and $0.21 respectively. H100 is 319% cheaper per token; H100 delivers 345% more tok/s/chip. (Numbers reflect the default 8k/1k Β· fp4 selection for this URL β table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H100:8609.6MI325X:4541.6 | H100:7305.9MI325X:3543.1 | H100:6412.7MI325X:1439.6 |
| Cost ($/M tok) | H100:$0.038MI325X:$0.067 | H100:$0.044MI325X:$0.086 | H100:$0.051MI325X:$0.212 |
| tok/s/MW | H100:6284399MI325X:2687353 | H100:5332791MI325X:2096534 | H100:4680812MI325X:851836 |
| Concurrency | H100:~34MI325X:~9 | H100:~32MI325X:~8 | H100:~32MI325X:~16 |
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.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom β’ Drag to pan β’ Double-click to reset β’ Click a point to pin tooltip
gpt-oss 120B 8K / 1K Total Tokens per $1 TCO vs. Interactivity