MiniMax M2.5/M2.7 — B200 vs H200
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H200 (NVIDIA Hopper) 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.
B200 posts 3245 tok/s/GPU for $0.17 per million tokens at 60 tok/s/user on MiniMax M2.5/M2.7; H200 posts 913 tok/s/GPU for $0.43. B200 is 157% cheaper per token; B200 delivers 255% more tok/s/GPU.
Throughput at 82 tok/s/user on MiniMax M2.5/M2.7: B200 hits 1515 tok/s/GPU, H200 hits 402. Per-million costs land at $0.36 and $0.95 respectively. B200 is 161% cheaper per token; B200 delivers 276% more tok/s/GPU.
B200 / H200 on MiniMax M2.5/M2.7 at 104 tok/s/user: 911 / 227 tok/s/GPU, $0.58 / $1.72 per million tokens. B200 is 198% cheaper per token; B200 delivers 302% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B200:3244.8H200:913.4 | B200:1515.0H200:402.4 | B200:910.8H200:226.8 |
| Cost ($/M tok) | B200:$0.167H200:$0.430 | B200:$0.362H200:$0.945 | B200:$0.580H200:$1.725 |
| tok/s/MW | B200:1495321H200:527982 | B200:698145H200:232615 | B200:419736H200:131109 |
| Concurrency | B200:~110H200:~62 | B200:~38H200:~20 | B200:~18H200:~9 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.