Qwen 3.5 397B-A17B — B200 vs B300
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and B300 (NVIDIA Blackwell) 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 76 tok/s/user on Qwen 3.5 397B-A17B: B200 hits 36600 tok/s/chip, B300 hits 45488. Per-million costs land at $0.01 and $0.01 respectively. B200 is 5% cheaper per token; B300 delivers 24% more tok/s/chip.
B200 / B300 on Qwen 3.5 397B-A17B at 145 tok/s/user: 29650 / 31100 tok/s/chip, $0.02 / $0.02 per million tokens. B200 is 25% cheaper per token; B300 delivers 5% more tok/s/chip.
Toward the upper edge of the 8–281 tok/s/user interactivity band, at 213 tok/s/user on Qwen 3.5 397B-A17B: B200 runs 14121 tok/s/chip at $0.03/M tokens, B300 runs 16709 at $0.04/M. B200 is 10% cheaper per token; B300 delivers 18% more tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:36600.2B300:45487.9 | B200:29650.1B300:31099.8 | B200:14121.1B300:16708.5 |
| Cost ($/M tok) | B200:$0.013B300:$0.014 | B200:$0.016B300:$0.020 | B200:$0.034B300:$0.038 |
| tok/s/MW | B200:21403608B300:23940979 | B200:17339264B300:16368341 | B200:8257934B300:8793955 |
| Concurrency | B200:~30B300:~40 | B200:~21B300:~22 | B200:~20B300:~11 |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
Total Tokens per $1 USD (Owning - Hyperscaler) vs. Interactivity
Qwen3.5 397B • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™
TCO $/chip/hr: H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68
Source: SemiAnalysis Market July 2026 Pricing Surveys & AI Cloud TCO Model
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate cost per million tokens per decode chip or per prefill chip, rather than per total chip count. This makes direct token cost comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate input throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct input throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate output throughput per decode chip or per prefill chip, rather than per total chip count. This makes direct output throughput comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate power per decode chip or per prefill chip, rather than per total chip count. This makes direct power comparison with aggregated configs not an apples-to-apples comparison.
Note: Disaggregated inference configurations (e.g., MoRI SGLang, Dynamo TRTLLM) calculate Joules per decode chip or per prefill chip, rather than per total chip count. This makes direct Joules per token comparison with aggregated configs not an apples-to-apples comparison.
No data available
Please change the model, sequence, precision, date range or chip selection.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip