Qwen 3.5 397B-A17B — B300 vs MI300X
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and MI300X (AMD CDNA 3) 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 4619 tok/s/chip for $0.14 per million tokens at 43 tok/s/user on Qwen 3.5 397B-A17B; MI300X posts 346 tok/s/chip for $0.76. B300 is 452% cheaper per token; B300 delivers 1236% more tok/s/chip.
Throughput at 53 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 3790 tok/s/chip, MI300X hits 190. Per-million costs land at $0.17 and $1.38 respectively. B300 is 730% cheaper per token; B300 delivers 1899% more tok/s/chip.
B300 / MI300X on Qwen 3.5 397B-A17B at 62 tok/s/user: 3207 / 123 tok/s/chip, $0.19 / $2.13 per million tokens. B300 is 997% cheaper per token; B300 delivers 2513% more tok/s/chip. (Numbers reflect the default 1k/1k · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:4618.6MI300X:345.8 | B300:3789.6MI300X:189.5 | B300:3206.9MI300X:122.7 |
| Cost ($/M tok) | B300:$0.137MI300X:$0.758 | B300:$0.167MI300X:$1.385 | B300:$0.195MI300X:$2.135 |
| tok/s/MW | B300:2430868MI300X:248748 | B300:1994539MI300X:136357 | B300:1687825MI300X:88306 |
| Concurrency | B300:~230MI300X:~34 | B300:~153MI300X:~15 | B300:~106MI300X:~8 |
Inference Performance
Inference performance metrics across different models, hardware configurations, and serving parameters.