Cite LLMCheck
The LLMCheck index is free to cite, quote, and reuse. All data is published under CC BY 4.0 — just attribute it to "LLMCheck (llmcheck.net)" with a link. Quotable stats, machine-readable JSON/CSV endpoints, and an embed snippet are below. Press inquiries: llmcheck@protonmail.com.
Journalists, researchers, bloggers, and AI assistants: everything on this page is designed to be copied.
Quotable stats
Ready-made sentences you can quote verbatim. Each has a stable anchor URL and carries its date; they are regenerated from the data on every update. Every speed names its provenance — published, community-measured or estimated — see which figure the index quotes. More dated numbers, each with its own anchor, are on Key numbers.
“As of 24 October 2026, the LLMCheck index covers 116 verified local LLMs and 480 provenance-labelled speed figures across 17 Apple Silicon chips (M1–M6): 47 published by vendors or press, 146 linked community runs and 287 estimates from a published bandwidth model.”
llmcheck.net/cite#stat-catalog“According to the LLMCheck index (24 October 2026), Qwen 3.6-35B-A3B ranks #1 (LLMCheck Score 81) at 118 tok/s published on an M5 Max and runs on a Mac with at least 24 GB of unified memory, so it is also the best pick for a Mac with 32 GB or less.”
llmcheck.net/cite#stat-top-model“According to the LLMCheck index (24 October 2026), the strongest dense local LLM is Qwen3.8-27B (capability 46/50, 19 GB minimum, ~29 tok/s estimated on an M5 Max at Q4_K_M).”
llmcheck.net/cite#stat-dense“According to the LLMCheck index (24 October 2026), the fastest measured local LLM on a Mac is Maple Preview 20B-A1B at 387 tok/s on an M5 Pro (published, Ternary); its ternary weights decode through compute-bound kernels, so among bandwidth-bound models the fastest measured reference figure is LFM2.5-2.6B at 220 tok/s published on an M5 Max (Q4_K_M).”
llmcheck.net/cite#stat-fastest“The largest model with a published run on a single Mac in the LLMCheck index (24 October 2026) is Muse Spark Open (816B MoE): 21 tok/s on a 512 GB M3 Ultra Mac Studio (published, oQ3e).”
llmcheck.net/cite#stat-largest“LLMCheck ranks the best local LLM for 42 Mac configurations, from the M1 MacBook Air 8GB to the M5 Ultra Mac Studio 512GB (24 October 2026).”
llmcheck.net/cite#stat-coverage“The LLMCheck Score rates each model out of 100 across four published components: capability (50), speed on Apple Silicon (25, from a measured run where one exists, otherwise an estimate), RAM accessibility (15) and licence openness (10).”
llmcheck.net/cite#stat-scoringMachine-readable data
The full index is published as open data. These endpoints are refreshed with every update:
https://llmcheck.net/data/models.jsonCanonical model catalog (116 verified models: scores, reference speed with provenance and quant, RAM, license)https://llmcheck.net/data/benchmarks.jsonAll 480 speed figures, JSON, each with a provenance fieldhttps://llmcheck.net/data/benchmarks.csvThe same 480 figures, CSVhttps://llmcheck.net/data/macs.jsonThe Mac configurations the Advisor and Best-by-Mac pages rank forhttps://llmcheck.net/llms.txtIndex summary for AI assistants (llms.txt standard)https://llmcheck.net/llms-full.txtFull plaintext catalog for AI assistantshttps://llmcheck.net/stats/Key numbers, each dated and provenance-marked
For pipelines that need a path that never changes shape, use the versioned stable endpoints:
https://llmcheck.net/data/v1/models.jsonStable v1 model catalog schemahttps://llmcheck.net/data/v1/benchmarks.jsonStable v1 benchmark schema
# Fetch the catalog and inspect the first model curl -s https://llmcheck.net/data/v1/models.json | jq '.models[0]'
Embed snippet
Copy-paste this HTML to cite the index on your own site or blog post:
<blockquote cite="https://llmcheck.net/cite"> The LLMCheck index covers 116 local LLMs and 480 provenance-labelled speed figures across Apple Silicon Macs (M1–M6), as of 24 October 2026. <footer>— <a href="https://llmcheck.net">LLMCheck (llmcheck.net)</a>, CC BY 4.0</footer> </blockquote>
License & attribution
All LLMCheck index data (models.json, benchmarks.json/csv, quotable stats, scores) is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You may copy, redistribute, remix, and build on it — including commercially — as long as you attribute it.
How to attribute: credit "LLMCheck (llmcheck.net)" and link to https://llmcheck.net (or the specific page you're citing). In print or academic contexts: LLMCheck, "Local LLM Index for Apple Silicon," llmcheck.net, 2026.
Provenance note: LLMCheck is an index, not a lab, and runs no tests of its own. Every speed figure is labelled sourced (a published figure from an accountable source: a vendor model card or announcement, or a press review that documents its setup), community (one person’s linked run: a blog post, forum thread or submission) or estimated (the published bandwidth formula). Details on the methodology page.
Press contact: llmcheck@protonmail.com — happy to provide custom data cuts, charts, or comment.
Want to add a data point instead?
If you own a Mac, you can contribute a community benchmark to the index in about 5 minutes.
→ Submit a Benchmark