Head-to-head AI inference benchmark comparison of GB200 NVL72 (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 →
Near the low end of the 53–180 tok/s/user interactivity band, at 84 tok/s/user on DeepSeekv4 Pro 0813 1.6T: GB200 NVL72 runs 5353 tok/s/chip at $0.10/M tokens, GB300 NVL72 runs 65802 at $0.01/M. GB300 NVL72 is 890% cheaper per token; GB300 NVL72 delivers 1129% more tok/s/chip.
Setting 116 tok/s/user as the target on DeepSeekv4 Pro 0813 1.6T, GB200 NVL72 produces 3599 tok/s/chip ($0.14 per million tokens) and GB300 NVL72 produces 25997 ($0.02). GB300 NVL72 is 482% cheaper per token; GB300 NVL72 delivers 622% more tok/s/chip.
At 148 tok/s/user interactivity on DeepSeekv4 Pro 0813 1.6T, GB200 NVL72 delivers 2605 tok/s/chip at $0.20 per million tokens; GB300 NVL72 delivers 12054 tok/s/chip at $0.05. GB300 NVL72 is 273% cheaper per token; GB300 NVL72 delivers 363% more tok/s/chip at this point. (Numbers reflect the default agentic-traces · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | GB200 NVL72:5353.0GB300 NVL72:65801.9 | GB200 NVL72:3598.5GB300 NVL72:25997.1 | GB200 NVL72:2605.3GB300 NVL72:12053.8 |
| Cost ($/M tok) | GB200 NVL72:$0.097GB300 NVL72:$0.010 | GB200 NVL72:$0.144GB300 NVL72:$0.025 | GB200 NVL72:$0.198GB300 NVL72:$0.053 |
| tok/s/MW | GB200 NVL72:2862581GB300 NVL72:31038631 | GB200 NVL72:1924356GB300 NVL72:12262777 | GB200 NVL72:1393201GB300 NVL72:5685752 |
| Concurrency | GB200 NVL72:~8GB300 NVL72:~690 | GB200 NVL72:~5GB300 NVL72:~186 | GB200 NVL72:~3GB300 NVL72:~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