TL;DR: A community repo now provides tested vLLM/llama.cpp configs for running large models on RTX 3090/4090/5090 cards.
Summary: The Club-3090 repository offers community recipes for serving LLMs on RTX 3090, 4090, and 5090 GPUs. It supports multiple inference engines (vLLM, llama.cpp, ik_llama) and is model-agnostic. Current configs cover Qwen3.6-27B, Qwen3.6-35B, Gemma 4 26B, and Gemma 4 31B on single or dual cards.
Why it matters: AI builders can skip trial-and-error when deploying small-to-mid-size LLMs on consumer hardware, getting production-like setups tuned for specific GPUs. Try these recipes if you're serving Qwen or Gemma models locally on 30/40/50-series cards.
Source: github_trending