Head-to-head AI inference benchmark comparison of B300 (NVIDIA Blackwell) and GB300 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.
B300 / GB300 NVL72 on GLM 5/5.1 at 59 tok/s/user: 2598 / 10515 tok/s/chip, $0.24 / $0.06 per million tokens. GB300 NVL72 is 296% cheaper per token; GB300 NVL72 delivers 305% more tok/s/chip.
Around the middle of the 30–147 tok/s/user interactivity band, at 88 tok/s/user on GLM 5/5.1: B300 runs 1755 tok/s/chip at $0.36/M tokens, GB300 NVL72 runs 7480 at $0.09/M. GB300 NVL72 is 317% cheaper per token; GB300 NVL72 delivers 326% more tok/s/chip.
Setting 118 tok/s/user as the target on GLM 5/5.1, B300 produces 1278 tok/s/chip ($0.49 per million tokens) and GB300 NVL72 produces 1761 ($0.36). GB300 NVL72 is 35% cheaper per token; GB300 NVL72 delivers 38% more tok/s/chip. (Numbers reflect the default 8k/1k · fp4 selection for this URL — table and chart below update if you change sequence, precision, or model in the controls.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | B300:2597.9GB300 NVL72:10515.5 | B300:1754.8GB300 NVL72:7480.0 | B300:1277.6GB300 NVL72:1761.2 |
| Cost ($/M tok) | B300:$0.242GB300 NVL72:$0.061 | B300:$0.358GB300 NVL72:$0.086 | B300:$0.491GB300 NVL72:$0.364 |
| tok/s/MW | B300:1367336GB300 NVL72:4960129 | B300:923579GB300 NVL72:3528310 | B300:672435GB300 NVL72:830752 |
| Concurrency | B300:~20GB300 NVL72:~1010 | B300:~9GB300 NVL72:~615 | B300:~5GB300 NVL72:~81 |
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