MiniMax M2.5/M2.7 — H200 vs MI300X
Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI300X (AMD CDNA 3) on MiniMax M2.5/M2.7. 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 27–95 tok/s/user interactivity band, at 44 tok/s/user on MiniMax M2.5/M2.7: H200 runs 1802 tok/s/GPU at $0.22/M tokens, MI300X runs 1249 at $0.25/M. H200 is 12% cheaper per token; H200 delivers 44% more tok/s/GPU.
Setting 61 tok/s/user as the target on MiniMax M2.5/M2.7, H200 produces 876 tok/s/GPU ($0.45 per million tokens) and MI300X produces 779 ($0.40). MI300X is 12% cheaper per token; H200 delivers 12% more tok/s/GPU.
At 78 tok/s/user interactivity on MiniMax M2.5/M2.7, H200 delivers 482 tok/s/GPU at $0.82 per million tokens; MI300X delivers 412 tok/s/GPU at $0.75. MI300X is 9% cheaper per token; H200 delivers 17% more tok/s/GPU at this point. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | H200:1801.9MI300X:1248.6 | H200:876.0MI300X:778.8 | H200:481.8MI300X:411.9 |
| Cost ($/M tok) | H200:$0.223MI300X:$0.250 | H200:$0.452MI300X:$0.402 | H200:$0.816MI300X:$0.747 |
| tok/s/MW | H200:1041557MI300X:697553 | H200:506371MI300X:435091 | H200:278486MI300X:230105 |
| Concurrency | H200:~80MI300X:~63 | H200:~63MI300X:~26 | H200:~21MI300X:~11 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.