Kimi K2.5/K2.6 1T — H200 vs MI355X Performance per Dollar
Cost per million tokens of H200 (NVIDIA Hopper) versus MI355X (AMD CDNA 4) on Kimi K2.5/K2.6 1T. Owning-hyperscaler TCO normalized by output tokens — performance per dollar across LLM workloads. Pick the more cost-efficient SKU at every target interactivity level. Use the chart controls below to switch sequences, precisions, and metrics — same interactions as the main inference chart.
At 36 tok/s/user on Kimi K2.5/K2.6 1T, H200 costs $0.83 per million tokens; MI355X costs $8.46. H200 is 916% more cost-efficient at this operating point.
H200 edges MI355X at 38 tok/s/user on Kimi K2.5/K2.6 1T — $0.89 per million tokens versus $8.88, a 897% cost-per-token gap.
Push Kimi K2.5/K2.6 1T to 41 tok/s/user and H200 lands at $0.98 per million tokens against MI355X's $9.45 — H200 pulls ahead by 866%. (Numbers reflect the default 1k/1k · int4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
GPU pricing (owning hyperscaler): H200 $1.41/GPU/hr · MI355X $1.48/GPU/hr. Source: SemiAnalysis Market August 2025 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | H200:$0.833MI355X:$8.464 | H200:$0.891MI355X:$8.883 | H200:$0.978MI355X:$9.448 |
| Concurrency | H200:~56MI355X:~6 | H200:~50MI355X:~5 | H200:~42MI355X:~4 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.