Head-to-head AI inference benchmark comparison of B200 (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 →
B200 posts 57589 tok/s/chip for $0.01 per million tokens at 130 tok/s/user on Qwen 3.5 397B-A17B; GB200 NVL72 posts 67848 tok/s/chip for $0.01. GB200 NVL72 is 10% cheaper per token; GB200 NVL72 delivers 18% more tok/s/chip.
Throughput at 217 tok/s/user on Qwen 3.5 397B-A17B: B200 hits 32482 tok/s/chip, GB200 NVL72 hits 33538. Per-million costs land at $0.01 and $0.02 respectively. B200 is 4% cheaper per token; GB200 NVL72 delivers 3% more tok/s/chip.
B200 / GB200 NVL72 on Qwen 3.5 397B-A17B at 303 tok/s/user: 13798 / 21950 tok/s/chip, $0.03 / $0.02 per million tokens. GB200 NVL72 is 48% cheaper per token; GB200 NVL72 delivers 59% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:57588.8GB200 NVL72:67847.8 | B200:32481.7GB200 NVL72:33538.0 | B200:13797.6GB200 NVL72:21950.1 |
| Cost ($/M tok) | B200:$0.008GB200 NVL72:$0.008 | B200:$0.015GB200 NVL72:$0.015 | B200:$0.035GB200 NVL72:$0.024 |
| tok/s/MW | B200:33677647GB200 NVL72:36282236 | B200:18995119GB200 NVL72:17934742 | B200:8068744GB200 NVL72:11738010 |
| Concurrency | B200:~22GB200 NVL72:~23 | B200:~23GB200 NVL72:~23 | B200:~10GB200 NVL72:~13 |
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.
TCO $/chip/hr:
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.
Matching measurements exist, but their chip series are hidden.
Shift+Scroll to zoom • Drag to pan • Double-click to reset • Click a point to pin tooltip
Qwen3.5 397B 8K / 1K Total Tokens per $1 TCO vs. Interactivity