Qwen 3.5 397B-A17B · GPU comparison

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 3055 tok/s/GPU for $0.21 per million tokens at 65 tok/s/user on Qwen 3.5 397B-A17B; H100 posts 427 tok/s/GPU for $0.86. B300 is 308% cheaper per token; B300 delivers 615% more tok/s/GPU.

Throughput at 100 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 1723 tok/s/GPU, H100 hits 311. Per-million costs land at $0.38 and $1.15 respectively. B300 is 202% cheaper per token; B300 delivers 455% more tok/s/GPU.

B300 / H100 on Qwen 3.5 397B-A17B at 136 tok/s/user: 1122 / 234 tok/s/GPU, $0.58 / $1.55 per million tokens. B300 is 168% cheaper per token; B300 delivers 380% 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)
B300:3055.1H100:427.2
B300:1723.4H100:310.5
B300:1121.7H100:233.9
Cost ($/M tok)
B300:$0.211H100:$0.861
B300:$0.381H100:$1.151
B300:$0.580H100:$1.553
tok/s/MW
B300:1407873H100:246955
B300:794206H100:179506
B300:516906H100:135210
Concurrency
B300:~96H100:~27
B300:~36H100:~13
B300:~18H100:~7

Inference Performance

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

Vendor:
Aggregation:
Spec Decoding: