Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and H100 (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 →
At 71 tok/s/user interactivity on Qwen 3.5 397B-A17B, B200 delivers 38013 tok/s/chip at $0.01 per million tokens; H100 delivers 6053 tok/s/chip at $0.05. B200 is 325% cheaper per token; B200 delivers 528% more tok/s/chip at this point.
B200 posts 33658 tok/s/chip for $0.01 per million tokens at 123 tok/s/user on Qwen 3.5 397B-A17B; H100 posts 3578 tok/s/chip for $0.09. B200 is 536% cheaper per token; B200 delivers 841% more tok/s/chip.
Throughput at 176 tok/s/user on Qwen 3.5 397B-A17B: B200 hits 21311 tok/s/chip, H100 hits 2528. Per-million costs land at $0.02 and $0.13 respectively. B200 is 470% cheaper per token; B200 delivers 743% more tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:38012.7H100:6052.8 | B200:33658.0H100:3577.7 | B200:21311.3H100:2527.6 |
| Cost ($/M tok) | B200:$0.013H100:$0.054 | B200:$0.014H100:$0.091 | B200:$0.023H100:$0.129 |
| tok/s/MW | B200:22229649H100:4418126 | B200:19683045H100:2611475 | B200:12462770H100:1844968 |
| Concurrency | B200:~32H100:~12 | B200:~24H100:~6 | B200:~16H100:~4 |
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