DeepSeekv4 Pro 0813 1.6T — B200 vs GB200 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on DeepSeekv4 Pro 0813 1.6T. 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 75 tok/s/user as the target on DeepSeekv4 Pro 0813 1.6T, B200 produces 9700 tok/s/chip ($0.05 per million tokens) and GB200 NVL72 produces 11946 ($0.04). GB200 NVL72 is 15% cheaper per token; GB200 NVL72 delivers 23% more tok/s/chip.
At 110 tok/s/user interactivity on DeepSeekv4 Pro 0813 1.6T, B200 delivers 5919 tok/s/chip at $0.08 per million tokens; GB200 NVL72 delivers 3795 tok/s/chip at $0.14. B200 is 68% cheaper per token; B200 delivers 56% more tok/s/chip at this point.
B200 posts 3373 tok/s/chip for $0.14 per million tokens at 145 tok/s/user on DeepSeekv4 Pro 0813 1.6T; GB200 NVL72 posts 2694 tok/s/chip for $0.19. B200 is 35% cheaper per token; B200 delivers 25% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:9700.1GB200 NVL72:11945.6 | B200:5918.6GB200 NVL72:3795.5 | B200:3373.3GB200 NVL72:2694.2 |
| Cost ($/M tok) | B200:$0.050GB200 NVL72:$0.043 | B200:$0.081GB200 NVL72:$0.136 | B200:$0.142GB200 NVL72:$0.192 |
| tok/s/MW | B200:5672546GB200 NVL72:6388027 | B200:3461145GB200 NVL72:2029668 | B200:1972669GB200 NVL72:1440737 |
| Concurrency | B200:~15GB200 NVL72:~38 | B200:~8GB200 NVL72:~5 | B200:~5GB200 NVL72:~3 |
Total Tokens per $1 USD (Owning - Neocloud Giant) vs. Interactivity
DeepSeek V4 Pro 0813 1.6T • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™
DeepSeek V4 Pro 0813 1.6T • 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
Interactivity is the rate at which a single user receives generated tokens while the model streams its answer — how quickly new words appear on screen. Higher values feel snappier; operators trade it against batch throughput.
How many total tokens (input + output) one US dollar of infrastructure spend buys, priced with the all-in hourly ownership cost of a Neocloud Giant operator. It is the reciprocal of cost per token, so higher means cheaper.
Formula: tok/$ = (total tok/s/chip × 3,600) ÷ all-in cost per chip-hour ($)