Stack Builder
Eight inputs — use case, budget, scale, privacy posture — and we compose the full rig: GPU + runtime + 1-3 model picks + first-run workflow + cost rollup + ready-to-paste install script. Three tiers side-by-side so the upgrade path stays visible.
Every recommendation references rule-based scoring; measured tok/s carries a confidence chip when surfaced. We don't invent numbers — when the data isn't there we say so.
Tell us about your build
URL updates as you change fields — share or bookmark a result. Showing the balanced default — change any field to refine.
Price figures are our planning estimates (launch MSRP, or a used-market estimate for older cards), not live prices. Check the current price before you buy.
Side-by-side: budget vs balanced vs stretch
one step down · your inputs · one step upYour inputs, our recommendation. Read the full card below.
One step up on budget. What you'd gain; what it costs.
Price figures are our planning estimates (launch MSRP, or a used-market estimate for older cards), not live prices. Check the current price before you buy. “Check price” links are affiliate links: we may earn a commission at no extra cost to you. How we make money.
Your recommended stack
full breakdown — read top to bottomNVIDIA GeForce RTX 3090 + Ollama + Qwen 3 14B
Not sure this card is enough for the stack? Rent the same GPU for an evening and run the model and runtime above first.
Price figures are our planning estimates (launch MSRP, or a used-market estimate for older cards), not live prices. Check the current price before you buy.
Affiliate and referral links: we may earn a commission at no extra cost to you. Prices and stock change, so check the listing. How we make money.
Default pick for most operators: one binary, automatic GPU detection, OpenAI-compatible HTTP API at `:11434`. Sufficient for solo + small-team workloads.
- ›Install:
curl -fsSL https://ollama.com/install.sh | sh - ›Pull a model:
ollama pull <model>:<tag> - ›HTTP API at
http://localhost:11434
- Qwen 3 14B14B paramsQ4_K_M~8.6 GB
Strongest general-purpose model at 14B in 2026. Multilingual tokenizer (1.7× more efficient on Turkish/Asian languages than Llama). Reasoning mode available.
CCommunity-reported·30-45 tok/s on 16GB VRAM - Phi-4 14B14B paramsQ4_K_M~8.5 GB
Microsoft's reasoning-focused 14B trained on heavy synthetic data. Beats Llama 3.1 8B on math/code benchmarks. Weaker creative writing.
EdEditorial·30-45 tok/s on 16GB VRAM
- Install Ollama on Linux.
- Pull the primary model:
ollama pull qwen3:14b - Verify it runs:
ollama run qwen3:14b— type a test prompt. - Connect a coding agent: install Cline (VS Code extension) or Aider (CLI) and point it at
http://localhost:11434(Ollama) or:8000(vLLM). - Verify the full loop end-to-end before adding observability, monitoring, or any second model.
Rent a GPU of the same class by the hour and run the runtime and model picks above on it first. If the model does not fit or feels slow there, it will not on the card you buy.
Referral links: RunLocalAI may earn a commission or account credits if you sign up and rent, at no extra cost to you. Hourly rates and which GPUs are listed change all the time, so check the provider's page before you start. How we make money.
Install script
copy and paste — this gets you to first token#!/usr/bin/env bash
# RunLocalAI stack installer — generated by /stack-builder
# Use case: coding
# Hardware: NVIDIA GeForce RTX 3090
# Runtime: Ollama
set -e
# 1. Install Ollama
curl -fsSL https://ollama.com/install.sh | sh
# 2. Pull the primary model
ollama pull qwen3:14b
# 3. Sanity check
ollama run qwen3:14b "Hello — respond in one short sentence to confirm you're running."
# 4. Verify the HTTP API
curl http://localhost:11434/api/tagsWhere to go from here
Just the hardware-pick question, with side-by-side compare, price/perf scatter, and score breakdown per dimension.
Reverse direction: I have this hardware — what fits? Use this to validate the recommendation against your actual rig.
Drill into the model picks: Q4_K_M vs Q5_K_M vs Q8 on your specific VRAM, with quality curve + VRAM fit visualization.
Tune every assumption: utilization, electricity rate, cloud equivalent rate, amortization horizon.
18 hand-curated stack recipes for specific outcomes (coding agent, offline RAG, dual-3090, Mac cluster, iPhone, etc.)
37 curated apps that plug into the runtime + model picks above: chat UIs, coding agents, RAG, voice, image, browser ext, mobile, editor plugins. Filter by your VRAM + privacy posture.