Open-source leaderboard whatmodel.sh shows the best local models your machine can run
jakecoolguy · reddit · 2026-10-12
A Redditor launched whatmodel.sh, an open-source AI leaderboard: enter your RAM and GPU to see which models your machine can run locally, with rankings, run commands, and score parity against API models.
- Each model gets a single score fitted from 19,731 published benchmark results, all linked to sources; the data, scoring code and bench script are open source (github.com/jakemanger/whatmodel).
- Key finding: a 32GB Mac can run a GPT-5-parity model fully locally and free (e.g. Qwen3 27B at 4-bit).
- Also works in the terminal via curl whatmodel.sh; contributions of real-world tok/s and new benchmark scores are welcome.
More from Infra
- Day 280/365 of GPU Programming: Diving into NCCL and How PyTorch Calls AllReduce — NandoDF · 2026-10-12
- FT: China rapidly builds data centers in energy-rich hinterland to power AI race — EleanorOlcott · 2026-10-12
- Lumentum Optics Sold Out Through Early 2029, Yet Only 44 of 149 AI Stocks Rose Last Week — demian_ai · 2026-10-12
- HF Engineer Runs pi-web on DGX Spark to Steer Local AI Agents From His Phone — lhoestq · 2026-10-12
- Running Qwen3.8-Flash-Next on six 3090s: 7 users at 262k context with a patched vLLM — flynth92 · 2026-10-12
- Grid becomes AI's new bottleneck: Entergy warned Amazon of peak-time cutoffs, AWS built Project Falcon — kimmonismus · 2026-10-12