GLM 5/5.1 · GPU comparison

GLM 5/5.1 — B300 vs GB200 NVL72

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/GPU at $0.25/M tokens, GB200 NVL72 runs 9553 at $0.06/M. GB200 NVL72 is 291% cheaper per token; GB200 NVL72 delivers 268% more tok/s/GPU.

Setting 88 tok/s/user as the target on GLM 5/5.1, B300 produces 1755 tok/s/GPU ($0.37 per million tokens) and GB200 NVL72 produces 6203 ($0.10). GB200 NVL72 is 272% cheaper per token; GB200 NVL72 delivers 253% more tok/s/GPU.

At 118 tok/s/user interactivity on GLM 5/5.1, B300 delivers 1278 tok/s/GPU at $0.50 per million tokens; GB200 NVL72 delivers 1574 tok/s/GPU at $0.39. GB200 NVL72 is 28% cheaper per token; GB200 NVL72 delivers 23% more tok/s/GPU 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.)

View performance-per-dollar view →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
B300:2597.9GB200 NVL72:9552.7
B300:1754.8GB200 NVL72:6202.9
B300:1277.6GB200 NVL72:1574.4
Cost ($/M tok)
B300:$0.252GB200 NVL72:$0.064
B300:$0.370GB200 NVL72:$0.099
B300:$0.505GB200 NVL72:$0.393
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

Inference Performance

Inference performance metrics across different models, hardware configurations, and serving parameters.

Vendor:
Aggregation:
Spec Decoding: