Qwen 3.5 397B-A17B · Chip comparison
Qwen 3.5 397B-A17B — B300 vs H200
Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and H200 (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.
AgentX replays real coding-agent sessions rather than fixed-length prompts, so context grows turn over turn and most of each request is served from cache instead of being recomputed. That turns the comparison into a systems question: KV transfer between nodes, prefix-aware routing, and cache capacity all move the curve alongside raw chip throughput. Fixed-sequence workloads stay the clean baseline for kernel and silicon performance, so the two scenarios answer different questions about the same hardware. Learn more about AgentX →
Throughput at 119 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 46,014 tok/s/chip, H200 hits 17,544. Costs per million tokens land at $0.014 and $0.019 respectively. B300 is 29% cheaper per token; B300 delivers 162% more tok/s/chip.
B300 / H200 on Qwen 3.5 397B-A17B at 199 tok/s/user: 20,738 / 10,796 tok/s/chip, $0.030 / $0.031 per million tokens. B300 is 4% cheaper per token; B300 delivers 92% more tok/s/chip.
Toward the upper edge of the 39–359 tok/s/user interactivity band, at 279 tok/s/user on Qwen 3.5 397B-A17B: B300 runs 10,038 tok/s/chip at $0.063 per million tokens, H200 runs 5,897 at $0.057. H200 is 8% cheaper per token; B300 delivers 70% more tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B30046,014H20017,544 | B30020,738H20010,796 | B30010,038H2005,897 |
| Cost ($/M tok) | B300$0.014H200$0.019 | B300$0.030H200$0.031 | B300$0.063H200$0.057 |
| tok/s/MW | B30024,217,888H20012,805,486 | B30010,914,720H2007,880,654 | B3005,283,408H2004,304,620 |
| Concurrency | B300~41H200~25 | B300~32H200~14 | B300~10H200~8 |