The local LLM tier list.
Community-voted rankings of the LLMs that run locally on Apple Silicon — auto-updated from Hugging Face, filtered to your Mac's unified memory. One Google account, one vote per model.
Quant playground.
Build a setup the way you'd build a quant: pick a Mac, pick a model, drop it to 3-bit, quantize the KV cache, switch on speculative decoding. Every number is computed live from a roofline model fitted to our own published benchmarks. No download, no waiting.
Pick a model
Speed and memory are computed for the settings above. An intelligence score marked * is estimated from scale, because Artificial Analysis has not scored that model.
Not yet ranked
Models with no published Artificial Analysis score. Real enough to run, not yet measured, so they aren't placed on a guess. Vote them up anyway: your votes are what will rank this board. Same hardware filters as above; rows hold their position while you vote.
| Model | Params | ~Size | Runs on | ↓ / mo | Tags | Your vote |
|---|