Small fast 'System 1' · Liquid AI
LFM2 1.2B
Hybrid convolution-attention model built for fast CPU decoding.
Datasheet
Maker | Liquid AI |
|---|---|
Parameters | 1.17 BVerifiedsafetensors.total from the HF API Source |
Licence | LFM Open License v1.0Verified |
Access | Public download |
Task (HF pipeline) | text-generation |
First published on HF | 2025-07-10 |
Last updated on HF | 2026-08-05 |
Weights in memory | 2.3 GB at 16-bit · 0.7 GB at ~4.85 bitsEstimateParameters × bits ÷ 8. Excludes context cache and runtime overhead; not every model has a 4-bit build. |
Benchmark
Not yet measured
BFCL V4, overall accuracy
No comparable public score for this benchmark yet. We will not estimate one.
The description summarises the model card at https://huggingface.co/LiquidAI/LFM2-1.2B. Parameters and licence are read from the Hugging Face API; the weights-in-memory line is our arithmetic, not a measurement.