LLM Inference Benchmark Explorer

GLM-4.7 on DGX B300 — FP8, vLLM, TP2 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelGLM-4.7
Parameters358B
Intelligence Index22.2
Agentic Index—
DeviceDGX B300
QuantizationFP8
Max C32+
Chat Capacity42
Agentic Capacity10
TP2
DP—
PP—
EnginevLLM
Speculative Decoding—
CTTFT (ms)TPS (tok/s)Status
15382.44PASS
27688.42PASS
46874.74PASS
87959.92PASS
1657145.13PASS
3264238.31PASS

In this measurement, GLM-4.7 (358B parameters), served in FP8 format with vLLM on DGX B300 (TP=2), reached a generation speed of 82.4 tok/s per request and a time to first token (TTFT) of 53 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is 42 users for chat use and 10 for agentic use. The realistic capacity will likely fall between these two values. In scenarios dominated by coding, tool use, long workflows and multi-agent use, capacity approaches the agentic estimate; where shorter interactions, standard conversations and lighter tasks dominate, it approaches the chat estimate.

Notes: 2 GPU

Intelligence Index and Agentic Index values are published by Artificial Analysis and are reproduced here with attribution.