Head-to-head AI inference benchmark comparison of GB300 NVL72 (NVIDIA Blackwell) and MI325X (AMD CDNA 3) 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 147 tok/s/user interactivity on Qwen 3.5 397B-A17B, GB300 NVL72 delivers 57797 tok/s/chip at $0.01 per million tokens; MI325X delivers 3631 tok/s/chip at $0.08. GB300 NVL72 is 658% cheaper per token; GB300 NVL72 delivers 1492% more tok/s/chip at this point.
GB300 NVL72 posts 24485 tok/s/chip for $0.03 per million tokens at 292 tok/s/user on Qwen 3.5 397B-A17B; MI325X posts 3631 tok/s/chip for $0.08. GB300 NVL72 is 221% cheaper per token; GB300 NVL72 delivers 574% more tok/s/chip.
Throughput at 437 tok/s/user on Qwen 3.5 397B-A17B: GB300 NVL72 hits 13807 tok/s/chip, MI325X hits 3631. Per-million costs land at $0.05 and $0.08 respectively. GB300 NVL72 is 81% cheaper per token; GB300 NVL72 delivers 280% more tok/s/chip. (Numbers reflect the default agentic-traces · fp8 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB300 NVL72:57796.8MI325X:3630.9 | GB300 NVL72:24484.6MI325X:3630.9 | GB300 NVL72:13806.7MI325X:3630.9 |
| Cost ($/M tok) | GB300 NVL72:$0.011MI325X:$0.084 | GB300 NVL72:$0.026MI325X:$0.084 | GB300 NVL72:$0.046MI325X:$0.084 |
| tok/s/MW | GB300 NVL72:27262631MI325X:2148478 | GB300 NVL72:11549340MI325X:2148478 | GB300 NVL72:6512618MI325X:2148478 |
| Concurrency | GB300 NVL72:~364MI325X:~8 | GB300 NVL72:~111MI325X:~8 | GB300 NVL72:~53MI325X:~8 |
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