Head-to-head AI inference benchmark comparison of H200 (NVIDIA Hopper) and TPU7x (Google Ironwood) 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 →
Setting 119 tok/s/user as the target on Qwen 3.5 397B-A17B, H200 produces 6555 tok/s/chip ($0.05 per million tokens) and TPU7x produces 39321 ($0.01). TPU7x is 611% cheaper per token; TPU7x delivers 500% more tok/s/chip.
At 157 tok/s/user interactivity on Qwen 3.5 397B-A17B, H200 delivers 5234 tok/s/chip at $0.06 per million tokens; TPU7x delivers 36758 tok/s/chip at $0.01. TPU7x is 732% cheaper per token; TPU7x delivers 602% more tok/s/chip at this point.
H200 posts 2803 tok/s/chip for $0.12 per million tokens at 195 tok/s/user on Qwen 3.5 397B-A17B; TPU7x posts 34772 tok/s/chip for $0.01. TPU7x is 1370% cheaper per token; TPU7x delivers 1141% more tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | H200:6554.9TPU7x:39321.0 | H200:5234.0TPU7x:36757.6 | H200:2802.7TPU7x:34772.0 |
| Cost ($/M tok) | H200:$0.052TPU7x:$0.007 | H200:$0.065TPU7x:$0.008 | H200:$0.121TPU7x:$0.008 |
| tok/s/MW | H200:4784564TPU7x:32577441 | H200:3820472TPU7x:30453653 | H200:2045764TPU7x:28808611 |
| Concurrency | H200:~11TPU7x:~27 | H200:~9TPU7x:~25 | H200:~3TPU7x:~23 |
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.
No measurements to plot for this selection. Review the benchmark controls above or adjust quick filters.
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