Qwen 3.5 397B-A17B — B300 vs H100
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H100 (NVIDIA Hopper) on Qwen 3.5 397B-A17B. 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 posts 3262 tok/s/GPU for $0.20 per million tokens at 61 tok/s/user on Qwen 3.5 397B-A17B; H100 posts 254 tok/s/GPU for $1.43. B300 is 621% cheaper per token; B300 delivers 1185% more tok/s/GPU.
Throughput at 78 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 2516 tok/s/GPU, H100 hits 196. Per-million costs land at $0.26 and $1.78 respectively. B300 is 589% cheaper per token; B300 delivers 1181% more tok/s/GPU.
B300 / H100 on Qwen 3.5 397B-A17B at 94 tok/s/user: 1913 / 155 tok/s/GPU, $0.34 / $2.38 per million tokens. B300 is 591% cheaper per token; B300 delivers 1136% more tok/s/GPU. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/gpu) | B300:3261.6H100:253.9 | B300:2515.6H100:196.4 | B300:1913.1H100:154.8 |
| Cost ($/M tok) | B300:$0.198H100:$1.429 | B300:$0.258H100:$1.779 | B300:$0.344H100:$2.381 |
| tok/s/MW | B300:1503023H100:146757 | B300:1159251H100:113529 | B300:881623H100:89489 |
| Concurrency | B300:~109H100:~17 | B300:~67H100:~10 | B300:~42H100:~7 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.