Kimi K3 2.8T — H200 vs MI355X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) 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 →
Near the low end of the 2–120 tok/s/user interactivity band, at 31 tok/s/user on Kimi K3 2.8T: H200 runs 70 tok/s/chip at $4.85/M tokens, MI355X runs 6889 at $0.06/M. MI355X is 7920% cheaper per token; MI355X delivers 9760% more tok/s/chip.
Setting 61 tok/s/user as the target on Kimi K3 2.8T, H200 produces 70 tok/s/chip ($4.85 per million tokens) and MI355X produces 3119 ($0.13). MI355X is 3531% cheaper per token; MI355X delivers 4364% more tok/s/chip.
At 90 tok/s/user interactivity on Kimi K3 2.8T, H200 delivers 70 tok/s/chip at $4.85 per million tokens; MI355X delivers 1932 tok/s/chip at $0.22. MI355X is 2150% cheaper per token; MI355X delivers 2666% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:69.9MI355X:6889.5 | H200:69.9MI355X:3119.0 | H200:69.9MI355X:1932.5 |
| Cost ($/M tok) | H200:$4.850MI355X:$0.060 | H200:$4.850MI355X:$0.134 | H200:$4.850MI355X:$0.216 |
| tok/s/MW | H200:51001MI355X:3296402 | H200:51001MI355X:1492361 | H200:51001MI355X:924639 |
| Concurrency | H200:~1MI355X:~22 | H200:~1MI355X:~6 | H200:~1MI355X:~3 |