LLM Inference Benchmark Explorer

Tencent-Hy4-preview on DGX B300 — BF16, vLLM, TP8, speculative decoding k=3 inference benchmark

Open this configuration in the LLM Inference Benchmark Explorer

ModelTencent-Hy4-preview
Parameters780B
Intelligence Index—
Agentic Index—
DeviceDGX B300
QuantizationBF16
Max C64+
Chat Capacity65
Agentic Capacity16
TP8
DP—
PP—
EnginevLLM
Speculative DecodingYes
CTTFT (ms)TPS (tok/s)Status
111065.66PASS
212361.16PASS
414655.73PASS
819046.80PASS
1626838.96PASS
3239830.76PASS
6447624.81PASS

In OpenZeka's measurement, Tencent-Hy4-preview (780B parameters), served in BF16 format with vLLM and speculative decoding on DGX B300 (TP=8), reached a generation speed of 65.7 tok/s per request and a time to first token (TTFT) of 110 ms with a single request.

Considering the speed targets and the available KV cache capacity, the estimated capacity is 65 users for chat use and 16 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: MTP k=3, 512K context. Model: tencent/Hy4-preview

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