MiniMax M2.5/M2.7 — B300 vs H100
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H100 (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.
B300 / H100 on MiniMax M2.5/M2.7 at 59 tok/s/user: 3463 / 614 tok/s/GPU, $0.19 / $0.58 per million tokens. B300 is 210% cheaper per token; B300 delivers 464% more tok/s/GPU.
Around the middle of the 40–117 tok/s/user interactivity band, at 78 tok/s/user on MiniMax M2.5/M2.7: B300 runs 1568 tok/s/GPU at $0.42/M tokens, H100 runs 378 at $0.94/M. B300 is 122% cheaper per token; B300 delivers 315% more tok/s/GPU.
Setting 98 tok/s/user as the target on MiniMax M2.5/M2.7, B300 produces 1067 tok/s/GPU ($0.59 per million tokens) and H100 produces 199 ($1.82). B300 is 209% cheaper per token; B300 delivers 435% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:3462.7H100:614.1 | B300:1568.3H100:378.2 | B300:1066.6H100:199.2 |
| Cost ($/M tok) | B300:$0.187H100:$0.580 | B300:$0.421H100:$0.937 | B300:$0.590H100:$1.824 |
| tok/s/MW | B300:1595712H100:354968 | B300:722720H100:218629 | B300:491505H100:115163 |
| Concurrency | B300:~157H100:~42 | B300:~47H100:~12 | B300:~22H100:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.