Meta's lfm2-24b-a2b is the largest on-device model yet, apache 2.0, built for always-on local…
meta's lfm2-24b-a2b is the largest on-device model yet, apache 2.0, built for always-on local agents that hold tool use and recovery across long tasks on one gpu. the interesting part isn't the size — it's that meta designed it around agent failures, not benchmarks.
local agents are now a hardware category.
Context
Liquid AI's blog of 24 February 2026 releases an early checkpoint of LFM2-24B-A2B, its largest LFM2 model, a sparse Mixture of Experts with 24 billion total and about 2 billion active parameters per token, designed to fit in 32GB of RAM on consumer laptops and desktops, open-weight on Hugging Face.
The owner is Liquid AI, and no text read attributes it to Meta. Largest means Liquid's own largest LFM2 model, not the largest on-device model overall. The Apache 2.0 claim was not verified, since the license file was not read and Liquid's models have used a custom license. One GPU and designed around agent failures were not found in the blog text. The post is dated February 2026, seven months before the note.
Related work
- LiquidAI/LFM2-24B-A2B (Hugging Face) ↗Snippet only.
- License file (Hugging Face) ↗Not read.
Watch next
- The license text.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 18 September 2026 at 12:48 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →