According to LLMCheck, the best way to choose a local LLM is comparing models across your specific priorities: speed, capability, RAM requirements, and license. Select 2–4 models below to see a detailed head-to-head comparison with LLMCheck Score breakdowns, or switch to Two Macs to see the same models’ speeds on two Mac configurations side by side.
Select at least 2 models to compare. Type a model name or company to filter.
Every speed is the LLMCheck index’s figure for that exact configuration, the same as its Best by Mac page. Ready-made pages for the most-asked pairs: Mac vs Mac.
⚖
Select models above and click Compare to see a side-by-side breakdown.
Frequently Asked Questions
How does the LLMCheck Model Comparator work?
Select 2 to 4 models from the searchable dropdown menus and click Compare. The tool generates a side-by-side table showing LLMCheck Score, score breakdown (Capability, Speed, Accessibility, License), tok/s on the leaderboard's reference chip (M5 Max, or M5 Ultra for models above 128 GB) with its provenance, minimum RAM, context window, architecture, license, and tags. The best value in each row is highlighted in green. You can share any comparison via a unique URL.
What metrics should I compare when choosing a local LLM for Mac?
The most important metrics are: LLMCheck Score (overall ranking out of 100), Capability Score (reasoning and benchmark quality, 0–50), tok/s (generation speed on the reference chip, marked published, community-measured or estimated), minimum RAM (must be less than your Mac's Unified Memory), and license (MIT and Apache 2.0 are the most permissive). MoE models offer large-model quality at small-model speed and RAM requirements.
Can I share my LLM comparison with others?
Yes. After running a comparison, click the Share This Comparison button to copy a unique URL. The URL contains model slugs as parameters so anyone opening the link sees the same side-by-side table instantly, with no account or setup required.
Can I compare two Macs instead of two models?
Yes. Switch to Two Macs, pick any two of the Mac configurations in the LLMCheck index, and the tool lists every model both can run with its decode speed on each Mac, side by side with its provenance, plus what only the bigger one fits, Apple's list price and price per GB. The figures are the same as the Best by Mac pages, and the most-asked pairs have their own Mac-vs-Mac pages.
Want the full ranked list?
See all 130 models sorted by LLMCheck Score, filterable by RAM tier, license, and architecture.