Qwen 3.5 397B-A17B · Chip comparison

Qwen 3.5 397B-A17B — GB200 NVL72 vs GB300 NVL72

Head-to-head AI inference benchmark comparison of GB200 NVL72 (NVIDIA Blackwell) and GB300 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

Throughput at 164 tok/s/user on Qwen 3.5 397B-A17B: GB200 NVL72 hits 47187 tok/s/chip, GB300 NVL72 hits 131676. Per-million costs land at $0.01 and $0.00 respectively. GB300 NVL72 is 125% cheaper per token; GB300 NVL72 delivers 179% more tok/s/chip.

GB200 NVL72 / GB300 NVL72 on Qwen 3.5 397B-A17B at 264 tok/s/user: 21005 / 87794 tok/s/chip, $0.02 / $0.01 per million tokens. GB300 NVL72 is 237% cheaper per token; GB300 NVL72 delivers 318% more tok/s/chip.

Toward the upper edge of the 64–465 tok/s/user interactivity band, at 365 tok/s/user on Qwen 3.5 397B-A17B: GB200 NVL72 runs 9447 tok/s/chip at $0.05/M tokens, GB300 NVL72 runs 46436 at $0.01/M. GB300 NVL72 is 296% cheaper per token; GB300 NVL72 delivers 392% 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)
GB200 NVL72:47186.9GB300 NVL72:131676.1
GB200 NVL72:21005.2GB300 NVL72:87794.1
GB200 NVL72:9446.6GB300 NVL72:46435.9
Cost ($/M tok)
GB200 NVL72:$0.011GB300 NVL72:$0.005
GB200 NVL72:$0.025GB300 NVL72:$0.007
GB200 NVL72:$0.055GB300 NVL72:$0.014
tok/s/MW
GB200 NVL72:25233631GB300 NVL72:62111365
GB200 NVL72:11232735GB300 NVL72:41412323
GB200 NVL72:5051667GB300 NVL72:21903728
Concurrency
GB200 NVL72:~15GB300 NVL72:~674
GB200 NVL72:~8GB300 NVL72:~423
GB200 NVL72:~7GB300 NVL72:~47

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