7 models from the LFM2 and LFM2.5 families at 0.64-0.69 bytes per weight, executable on CPU by a C runtime whose entire dependency list is libc, libm and libgomp.
| model | package | RAM to run | compression | perplexity | rel L2 |
|---|---|---|---|---|---|
| LFM2.5-230M | 161 MB | 198 MB | 2.89x | +8.43% | 0.0485 |
| LFM2-350M | 241 MB | 294 MB | 2.96x | +1.10%* | 0.0533 |
| LFM2-700M | 495 MB | 573 MB | 3.01x | +12.51% | 0.0547 |
| LFM2-1.2B | 774 MB | 873 MB | 3.03x | +9.91% | 0.0553 |
| LFM2.5-2.6B | 1.77 GB | 2.00 GB | 3.05x | -2.65% | 0.0544 |
| LFM2.5-8B-A1B | 5.49 GB | 6.33 GB | 3.09x | +2.58%* | 0.0548 |
| LFM2-24B-A2B | 15.37 GB | 17.84 GB | 3.10x | none detected | 0.0561 |
* resolvable individually but not after correcting for 7 simultaneous tests, which needs |t| > 2.89.
Relative L2 is how faithfully the weights are reproduced — 0.049–0.056 across every model here, holding steady across four tokenizers, dense and Mixture-of-Experts architectures, and a 100× range in parameter count. It is the figure that describes the compression.
Perplexity is how a model scores on a corpus, measured against the bf16 checkpoint on identical tokens and clustered by 512-token window. 1 of these show no detectable cost. It varies with each model’s sensitivity and with how well general web text matches its domain — two models here compressed to the same 0.055 rel L2 measure +13.4% and −3.2%.
git clone https://github.com/EntroMorphic/cassie && cd cassie cmake -S src/c -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build -j huggingface-cli download ZTFlynn/LFM2-350M-Cascadia-ternary3 --local-dir ./pkg ./build/cascadia_generate ./pkg 512 --chat "Explain gradient descent."
Deeply inspired by Magneato/deepseek-r1-qwen-7b-lutc, which demonstrated LUT-cascade compression at 5.45 bits per weight. Base models by Liquid AI, used under the LFM Open License. Method and measurements: github.com/EntroMorphic/cassie.