# noonghunna/club-3090

Community recipes for serving LLMs on RTX 3090/4090/5090 CUDA gpus. Multi-engine (vLLM, llama.cpp, ik_llama) and model-agnostic. Currently shipping Qwen3.6-27B Qwen3.6 35B Gemma 4 26B Gemma 4 31B configs for 1× and 2× cards.

Repository: https://github.com/noonghunna/club-3090
Canonical: https://ross.abutalabs.com/products/club-3090
Language: Python
License: Apache-2.0
License Family: permissive
Last push: 2026-08-26T18:17:16+00:00

## Health v2 (maintenance only)
Score: 79/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 93, longevity 9
- inputs: {"age_days": 127, "days_push": 7, "days_rel": 51, "gap_med": 0, "n_releases_24m": 32}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2101, forks 127 (observed 2026-08-28T04:06:13.448995+00:00)

## What it is
A community-maintained collection of recipes, configs, and scripts for serving large language models locally on consumer NVIDIA GPUs (RTX 3090/4090/5090). It is multi-engine (vLLM, llama.cpp, ik_llama) and model-agnostic, shipping curated launch scripts, hardware-aware pickers, and benchmarks for single- and dual-GPU setups.

## Use cases
- serve llms locally on rtx 3090
- run qwen or gemma on two consumer gpus
- find working vllm configs for 24gb cards
- self-host an llm backend in a homelab
- benchmark llama.cpp vs vllm on consumer gpus
- set up local llm with open webui and image generation

## When to choose
- You own one or two RTX 3090/4090/5090 cards and want proven serving configs
- You want a scripted, hardware-aware setup for local LLM inference on Linux/macOS or WSL2
- You want curated per-model defaults with VRAM budgeting and benchmarks

## When to avoid
- You need production multi-node or datacenter GPU serving
- You run native Windows without WSL2 (tooling is unsupported there)
- You need a general-purpose inference library rather than curated configs and scripts

## Facets
- artifact type: infra-config
- maturity: active
- function: llm-inference, configuration-management, deployment, benchmarking, developer-tools
- domain: large-language-models, self-hosted, gpu-computing, developer-tools, hardware
- platform: self-hosted, cli
- tags: consumer-gpus, rtx-3090, rtx-4090, rtx-5090, vllm, llama-cpp, local-llm, homelab, model-serving, recipes, qwen, gemma, linux, macos, gpu, docker

## Member repositories
- noonghunna/club-3090 (main) score 79

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:13.448995+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:54:18.624859+00:00, confidence not recorded.
  - readme: https://github.com/noonghunna/club-3090 (fetched 2026-08-28T04:06:13.448995+00:00, sha 0df70840a065)
- Data as of 2026-08-30T08:39:29.467469+00:00.