Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and MI355X (AMD CDNA 4) on Kimi K3 2.8T. 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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Throughput at 45 tok/s/user on Kimi K3 2.8T: B200 hits 4996 tok/s/chip, MI355X hits 7636. Per-million costs land at $0.10 and $0.05 respectively. MI355X is 76% cheaper per token; MI355X delivers 53% more tok/s/chip.
B200 / MI355X on Kimi K3 2.8T at 80 tok/s/user: 2910 / 3602 tok/s/chip, $0.17 / $0.12 per million tokens. MI355X is 43% cheaper per token; MI355X delivers 24% more tok/s/chip.
Toward the upper edge of the 9–151 tok/s/user interactivity band, at 116 tok/s/user on Kimi K3 2.8T: B200 runs 1834 tok/s/chip at $0.26/M tokens, MI355X runs 2827 at $0.15/M. MI355X is 78% cheaper per token; MI355X delivers 54% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:4996.1MI355X:7635.6 | B200:2910.0MI355X:3601.9 | B200:1834.0MI355X:2826.8 |
| Cost ($/M tok) | B200:$0.096MI355X:$0.055 | B200:$0.165MI355X:$0.116 | B200:$0.262MI355X:$0.147 |
| tok/s/MW | B200:2921721MI355X:3653384 | B200:1701766MI355X:1723392 | B200:1072528MI355X:1352522 |
| Concurrency | B200:~18MI355X:~15 | B200:~9MI355X:~6 | B200:~5MI355X:~4 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
