Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB200 NVL72 (NVIDIA Blackwell) on GLM 5/5.1. 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.
Near the low end of the 30–147 tok/s/user interactivity band, at 59 tok/s/user on GLM 5/5.1: B300 runs 2598 tok/s/chip at $0.24/M tokens, GB200 NVL72 runs 9553 at $0.05/M. GB200 NVL72 is 347% cheaper per token; GB200 NVL72 delivers 268% more tok/s/chip.
Setting 88 tok/s/user as the target on GLM 5/5.1, B300 produces 1755 tok/s/chip ($0.36 per million tokens) and GB200 NVL72 produces 6203 ($0.08). GB200 NVL72 is 330% cheaper per token; GB200 NVL72 delivers 253% more tok/s/chip.
At 118 tok/s/user interactivity on GLM 5/5.1, B300 delivers 1278 tok/s/chip at $0.49 per million tokens; GB200 NVL72 delivers 1574 tok/s/chip at $0.33. GB200 NVL72 is 50% cheaper per token; GB200 NVL72 delivers 23% more tok/s/chip at this point. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:2597.9GB200 NVL72:9552.7 | B300:1754.8GB200 NVL72:6202.9 | B300:1277.6GB200 NVL72:1574.4 |
| Cost ($/M tok) | B300:$0.242GB200 NVL72:$0.054 | B300:$0.358GB200 NVL72:$0.083 | B300:$0.491GB200 NVL72:$0.328 |
| tok/s/MW | B300:1367336GB200 NVL72:5108387 | B300:923579GB200 NVL72:3317076 | B300:672435GB200 NVL72:841925 |
| Concurrency | B300:~20GB200 NVL72:~916 | B300:~9GB200 NVL72:~661 | B300:~5GB200 NVL72:~75 |
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
GLM5/5.1 744B 8K / 1K Total Tokens per $1 TCO vs. Interactivity