LLM Inference Benchmark Explorer

MiMo-V2.6-Pro-RL on DGX B300 — FP4, vLLM, TP4, speculative decoding k=7 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelMiMo-V2.6-Pro-RL
Parameters1.02T
Intelligence Index46.3
Agentic Index—
DeviceDGX B300
QuantizationFP4
Max C64+
Chat Capacity256+
Agentic Capacity96+
TP4
DP—
PP—
EnginevLLM
Speculative DecodingYes
CTTFT (ms)TPS (tok/s)Status
146333.47PASS
257271.25PASS
468226.55PASS
890172.29PASS
16117137.18PASS
32166110.04PASS
6416889.53PASS

In OpenZeka's measurement, MiMo-V2.6-Pro-RL (1.02T parameters), served in FP4 format with vLLM and speculative decoding on DGX B300 (TP=4), reached a generation speed of 333.5 tok/s per request and a time to first token (TTFT) of 46 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is at least 256 users for chat use and at least 96 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.

“At least” means the configuration met the speed targets even at the highest load tested, so the real capacity may be higher.

Notes: DFlash speculative decoding, 1M context. Model: XiaomiMiMo/MiMo-V2.6-Pro-RL

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