Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on DeepSeek R1. 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.
B300 / GB200 NVL72 on DeepSeek R1 at 93 tok/s/user: 8128 / 11532 tok/s/chip, $0.08 / $0.04 per million tokens. GB200 NVL72 is 72% cheaper per token; GB200 NVL72 delivers 42% more tok/s/chip.
Around the middle of the 24–299 tok/s/user interactivity band, at 162 tok/s/user on DeepSeek R1: B300 runs 1801 tok/s/chip at $0.35/M tokens, GB200 NVL72 runs 4192 at $0.12/M. GB200 NVL72 is 183% cheaper per token; GB200 NVL72 delivers 133% more tok/s/chip.
Setting 231 tok/s/user as the target on DeepSeek R1, B300 produces 1055 tok/s/chip ($0.60 per million tokens) and GB200 NVL72 produces 1012 ($0.51). GB200 NVL72 is 17% cheaper per token; B300 delivers 4% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:8127.6GB200 NVL72:11531.8 | B300:1800.8GB200 NVL72:4192.5 | B300:1054.8GB200 NVL72:1012.2 |
| Cost ($/M tok) | B300:$0.077GB200 NVL72:$0.045 | B300:$0.349GB200 NVL72:$0.123 | B300:$0.595GB200 NVL72:$0.510 |
| tok/s/MW | B300:4277680GB200 NVL72:6166725 | B300:947790GB200 NVL72:2241975 | B300:555139GB200 NVL72:541259 |
| Concurrency | B300:~194GB200 NVL72:~913 | B300:~45GB200 NVL72:~178 | B300:~33GB200 NVL72:~19 |
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
DeepSeek R1 0528 671B 8K / 1K Total Tokens per $1 TCO vs. Interactivity