Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (AMD CDNA 3) on Kimi K2.5/K2.6/K2.7-Code 1T. 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.
Near the low end of the 17–43 tok/s/user interactivity band, at 23 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T: H200 runs 1074 tok/s/chip at $0.32/M tokens, MI300X runs 386 at $0.68/M. H200 is 117% cheaper per token; H200 delivers 178% more tok/s/chip.
Setting 30 tok/s/user as the target on Kimi K2.5/K2.6/K2.7-Code 1T, H200 produces 974 tok/s/chip ($0.35 per million tokens) and MI300X produces 289 ($0.91). H200 is 163% cheaper per token; H200 delivers 237% more tok/s/chip.
At 37 tok/s/user interactivity on Kimi K2.5/K2.6/K2.7-Code 1T, H200 delivers 896 tok/s/chip at $0.38 per million tokens; MI300X delivers 224 tok/s/chip at $1.18. H200 is 212% cheaper per token; H200 delivers 301% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:1074.0MI300X:386.3 | H200:973.5MI300X:288.7 | H200:896.1MI300X:223.5 |
| Cost ($/M tok) | H200:$0.316MI300X:$0.683 | H200:$0.348MI300X:$0.914 | H200:$0.378MI300X:$1.181 |
| tok/s/MW | H200:783946MI300X:277903 | H200:710611MI300X:207727 | H200:654072MI300X:160804 |
| Concurrency | H200:~45MI300X:~16 | H200:~30MI300X:~9 | H200:~22MI300X:~5 |
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
Kimi K2.5/2.6/2.7-Code 1T 8K / 1K Total Tokens per $1 TCO vs. Interactivity