MiniMax M2.5/M2.7 · GPU comparison

MiniMax M2.5/M2.7 — H200 vs MI355X

Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and MI355X (AMD CDNA 4) 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.

Throughput at 58 tok/s/user on MiniMax M2.5/M2.7: H200 hits 983 tok/s/GPU, MI355X hits 1708. Per-million costs land at $0.39 and $0.25 respectively. MI355X is 60% cheaper per token; MI355X delivers 74% more tok/s/GPU.

H200 / MI355X on MiniMax M2.5/M2.7 at 78 tok/s/user: 438 / 828 tok/s/GPU, $0.86 / $0.48 per million tokens. MI355X is 78% cheaper per token; MI355X delivers 89% more tok/s/GPU.

Toward the upper edge of the 39–117 tok/s/user interactivity band, at 98 tok/s/user on MiniMax M2.5/M2.7: H200 runs 270 tok/s/GPU at $1.45/M tokens, MI355X runs 475 at $0.84/M. MI355X is 72% cheaper per token; MI355X delivers 76% more tok/s/GPU. (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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
H200:982.6MI355X:1707.8
H200:438.3MI355X:828.3
H200:269.6MI355X:475.2
Cost ($/M tok)
H200:$0.394MI355X:$0.246
H200:$0.863MI355X:$0.485
H200:$1.453MI355X:$0.842
tok/s/MW
H200:567955MI355X:644455
H200:253368MI355X:312550
H200:155843MI355X:179319
Concurrency
H200:~70MI355X:~61
H200:~23MI355X:~15
H200:~11MI355X:~8

Inference Performance

Inference performance metrics across different models, hardware configurations, and serving parameters.