onprem-llm-benchmark

B待确认Benchmark

性能/成本推理速度开放cc-by-4.0

发布方：Scherm-AI

热度17.1▼ 0.2

Hugging Face

下载量 · 30天

4,044

Hugging Face

GitHub Stars

—

代码仓库

论文被引

—

Semantic Scholar

跑分模型 · 30天

—

Leaderboard results

简介

Scherm On-Premise LLM Inference Benchmark — v0.8 Throughput, latency and resource benchmarks for on-premise LLM inference across a wide range of GPUs — datacenter (B200, L40S), Blackwell workstation (RTX PRO 6000), consumer (RTX 4090/4070), unified-memory (DGX Spark / GB10) and legacy/AMD — covering the full stack of a production assistant: chat serving, embeddings, reranking, vision, OCR and speech-to-text. Engine: vLLM (mostly the NGC 26.03 image, engine v0.17.1; Blackwell… See the full description on the dataset page: https://huggingface.co/datasets/Scherm-AI/onprem-llm-benchmark.