Kimi K2.5/K2.6/K2.7-Code 1T — B200 vs MI355X Performance per Dollar
Cost per million tokens of B200 (NVIDIA Blackwell) versus MI355X (AMD CDNA 4) on Kimi K2.5/K2.6/K2.7-Code 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.
On Kimi K2.5/K2.6/K2.7-Code 1T at 38 tok/s/user, the per-million math comes out to $0.04 for B200 and $0.10 for MI355X; B200 delivers 161% more output per dollar.
At 67 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T, B200 costs $0.13 per million tokens; MI355X costs $0.17. B200 is 27% more cost-efficient at this operating point.
MI355X edges B200 at 95 tok/s/user on Kimi K2.5/K2.6/K2.7-Code 1T — $0.24 per million tokens versus $0.29, a 23% cost-per-token gap. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Chip pricing (owning hyperscaler): B200 $1.73/chip/hr · MI355X $1.50/chip/hr. Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Dollar per Million Tokens | B200:$0.040MI355X:$0.105 | B200:$0.132MI355X:$0.167 | B200:$0.292MI355X:$0.238 |
| Concurrency | B200:~736MI355X:~102 | B200:~178MI355X:~17 | B200:~9MI355X:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.