DeepSeekv4 Pro 0813 1.6T — B200 vs GB300 NVL72
Head-to-head AI inference benchmark comparison of B200 (NVIDIA Blackwell) and GB300 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 →
Throughput at 72 tok/s/user on DeepSeekv4 Pro 0813 1.6T: B200 hits 8200 tok/s/chip, GB300 NVL72 hits 82164. Per-million costs land at $0.06 and $0.01 respectively. GB300 NVL72 is 650% cheaper per token; GB300 NVL72 delivers 902% more tok/s/chip.
B200 / GB300 NVL72 on DeepSeekv4 Pro 0813 1.6T at 109 tok/s/user: 5845 / 6564 tok/s/chip, $0.08 / $0.10 per million tokens. B200 is 19% cheaper per token; GB300 NVL72 delivers 12% more tok/s/chip.
Toward the upper edge of the 34–185 tok/s/user interactivity band, at 147 tok/s/user on DeepSeekv4 Pro 0813 1.6T: B200 runs 3037 tok/s/chip at $0.16/M tokens, GB300 NVL72 runs 3626 at $0.18/M. B200 is 12% cheaper per token; GB300 NVL72 delivers 19% more tok/s/chip. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B200:8200.2GB300 NVL72:82163.8 | B200:5845.4GB300 NVL72:6564.4 | B200:3036.7GB300 NVL72:3625.9 |
| Cost ($/M tok) | B200:$0.059GB300 NVL72:$0.008 | B200:$0.082GB300 NVL72:$0.098 | B200:$0.158GB300 NVL72:$0.177 |
| tok/s/MW | B200:4795454GB300 NVL72:38756496 | B200:3418351GB300 NVL72:3096415 | B200:1775842GB300 NVL72:1710337 |
| Concurrency | B200:~14GB300 NVL72:~993 | B200:~8GB300 NVL72:~5 | B200:~4GB300 NVL72:~2 |
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 - Hyperscaler) vs. Interactivity
DeepSeek V4 Pro 1.6T • FP4 • 8K / 1K • Source: SemiAnalysis InferenceX™
TCO $/chip/hr: H100: 1.17H200: 1.22B200: 1.73B300: 2.26GB200: 1.86GB300: 2.31MI300X: 0.95MI325X: 1.1MI355X: 1.5RTX6000PRO: 0.68
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