InferenceX
中文
Qwen 3.5 397B-A17B · Chip comparison

Qwen 3.5 397B-A17B — B300 vs GB200 NVL72

Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB200 NVL72 (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 →

At 143 tok/s/user interactivity on Qwen 3.5 397B-A17B, B300 delivers 43855 tok/s/chip at $0.01 per million tokens; GB200 NVL72 delivers 58915 tok/s/chip at $0.01. GB200 NVL72 is 63% cheaper per token; GB200 NVL72 delivers 34% more tok/s/chip at this point.

B300 posts 29074 tok/s/chip for $0.02 per million tokens at 221 tok/s/user on Qwen 3.5 397B-A17B; GB200 NVL72 posts 28755 tok/s/chip for $0.02. GB200 NVL72 is 20% cheaper per token; B300 delivers 1% more tok/s/chip.

Throughput at 300 tok/s/user on Qwen 3.5 397B-A17B: B300 hits 11944 tok/s/chip, GB200 NVL72 hits 13538. Per-million costs land at $0.05 and $0.04 respectively. GB200 NVL72 is 38% cheaper per token; GB200 NVL72 delivers 13% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 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/chip)
B300:43854.8GB200 NVL72:58914.5
B300:29074.4GB200 NVL72:28754.5
B300:11943.5GB200 NVL72:13538.5
Cost ($/M tok)
B300:$0.014GB200 NVL72:$0.009
B300:$0.022GB200 NVL72:$0.018
B300:$0.053GB200 NVL72:$0.038
tok/s/MW
B300:23081462GB200 NVL72:31505103
B300:15302342GB200 NVL72:15376741
B300:6286063GB200 NVL72:7239823
Concurrency
B300:~18GB200 NVL72:~21
B300:~19GB200 NVL72:~20
B300:~5GB200 NVL72:~5

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 - Neocloud Giant) vs. Interactivity

Qwen3.5 397B • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™

TCO $/chip/hr: VR200: 3.61H100: 1.55H200: 1.59B200: 2.07B300: 2.52GB200: 2.26GB300: 2.79MI300X: 1.16MI325X: 1.32MI355X: 2.09RTX6000PRO: 0.75Jalapeño (Teacup): 1.56

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