# Local AI PC build recipes | What Can I Run AI?

> Exact buy lists for running local LLMs: dual 3090 for 70B, 24 GB budget builds, Mac Studio vs. GPUs, with totals and fit checks.

Source: https://whatcanirunai.vercel.app/builds/

# Build recipes: goal to exact buy list

Pick a goal. Each recipe lists every part with a dated price, checks the PSU, case and slots, and estimates speed. Want to change a part? Open it in the [planner](https://whatcanirunai.vercel.app/planner/). Weighing two or three? [Compare builds side by side](https://whatcanirunai.vercel.app/builds/compare/).

[Dual used RTX 3090 build (48 GB) for 32B to 70B models48.0 GB usable936 GB/s2× GPUTarget: DeepSeek-R1-Distill-Qwen-32B**$7,216**7 parts →](https://whatcanirunai.vercel.app/builds/dual-used-rtx-3090-70b/)[Cheapest PC build that runs Qwen3.8-27B at Q424.0 GB usable960 GB/sTarget: Qwen3.8-27B**$2,427**7 parts →](https://whatcanirunai.vercel.app/builds/cheapest-build-qwen3-8-27b/)[Mac Studio M5 Ultra vs. dual RTX 3090 for local LLMs192.0 GB usable1,200 GB/sTarget: Llama-3.3-70B-Instruct**$9,499**1 parts →](https://whatcanirunai.vercel.app/builds/mac-studio-vs-dual-rtx-3090/)[Budget 24 GB VRAM build24.0 GB usable960 GB/sTarget: Qwen3-32B**$2,427**7 parts →](https://whatcanirunai.vercel.app/builds/24gb-budget-build/)[16 GB starter build (RTX 4060 Ti 16GB)16.0 GB usable288 GB/sTarget: Qwen3-14B**$1,887**7 parts →](https://whatcanirunai.vercel.app/builds/16gb-starter-build/)[Cheapest 12 GB build: used RTX 306012.0 GB usable360 GB/sTarget: Qwen3-8B**$1,778**7 parts →](https://whatcanirunai.vercel.app/builds/12gb-used-rtx-3060-build/)[RTX 5090 single-GPU build (32 GB)32.0 GB usable1,792 GB/sTarget: Qwen3-32B**$4,533**7 parts →](https://whatcanirunai.vercel.app/builds/rtx-5090-32gb-build/)[RTX 4090 build for 24 GB models24.0 GB usable1,008 GB/sTarget: gemma-3-27b-it**$4,023**7 parts →](https://whatcanirunai.vercel.app/builds/rtx-4090-build/)[New-parts budget build: RTX 5070 12GB12.0 GB usable672 GB/sTarget: Qwen3-8B**$1,937**7 parts →](https://whatcanirunai.vercel.app/builds/rtx-5070-new-budget-build/)[Dual RX 7900 XTX build (48 GB VRAM)48.0 GB usable960 GB/s2× GPUTarget: Qwen2.5-Coder-32B-Instruct**$5,134**7 parts →](https://whatcanirunai.vercel.app/builds/dual-rx-7900-xtx-48gb-build/)[Local coding agent build (Qwen3-Coder-30B-A3B)24.0 GB usable1,008 GB/sTarget: Qwen3-Coder-30B-A3B-Instruct**$4,023**7 parts →](https://whatcanirunai.vercel.app/builds/coding-agent-build/)[What runs gpt-oss-120b locally: Framework Desktop 128GB96.0 GB usable256 GB/sTarget: gpt-oss-120b**$3,449**1 parts →](https://whatcanirunai.vercel.app/builds/gpt-oss-120b-machine/)[DGX Spark vs. Framework Desktop for local AI96.0 GB usable273 GB/sTarget: gpt-oss-120b**$6,950**1 parts →](https://whatcanirunai.vercel.app/builds/dgx-spark-vs-framework-desktop/)[Quiet starter: Mac mini M6 32GB24.0 GB usable170 GB/sTarget: Qwen3-14B**$1,799**1 parts →](https://whatcanirunai.vercel.app/builds/mac-mini-m6-32-quiet-build/)[Two used RTX 3060 12GB cards for 24 GB24.0 GB usable360 GB/s2× GPUTarget: Qwen3-32B**$2,208**7 parts →](https://whatcanirunai.vercel.app/builds/dual-used-rtx-3060-24gb/)

Prices: Corsair store (PSUs, cases, RAM) and Newegg (CPUs, boards, SSDs) on 2026-10-23, GPUs at list price unless marked used. DDR5 prices are unusually high right now, and they dominate several totals.

---
Agent guide: https://whatcanirunai.vercel.app/llms.txt · JSON API: https://whatcanirunai.vercel.app/developers/