LLM Inference Benchmark Explorer

Kimi K3 on DGX B300 — MXFP4, SGLang, TP8 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelKimi K3
Parameters2.78T
Intelligence Index43.6
Agentic Index50
DeviceDGX B300
QuantizationMXFP4
Max C64+
Chat Capacity150
Agentic Capacity50
TP8
DP—
PP—
EngineSGLang
Speculative Decoding—
CTTFT (ms)TPS (tok/s)Status
140083.40PASS
277263.79PASS
481159.33PASS
878352.88PASS
1683443.54PASS
3291533.55PASS
6494829.11PASS

In OpenZeka's measurement, Kimi K3 (2.78T parameters), served in MXFP4 format with SGLang on DGX B300 (TP=8), reached a generation speed of 83.4 tok/s per request and a time to first token (TTFT) of 400 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is 150 users for chat use and 50 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: GMU 0.85

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