Liquid Foundation Models

Fastest on-device foundation models.

Purpose-built for edge AI, our hybrid models run entirely on-device and shine at agentic tasks, data extraction, RAG, creative writing, and multi-turn conversations.

Edge

LFM2-350M

Multilingual
Fine-tuning support
On-device deployment

Learn more

Edge

LFM2-700M

Multilingual
Fine-tuning support
On-device deployment

Learn more

Edge

LFM2-1.2B

Multilingual
Fine-tuning support
On-device deployment

Learn more

Start building with LFMs today

Try LEAP, our developer-first platform for AI on the edge, proudly OS- and model-agnostic.

Try Leap

Download Apollo

State-of-the-art performance at every scale*.

Benchmarks

LFM2-350M

LFM2-700M

LFM2-1.2B

MMLU

(5-shot)

43.43

49.9

55.23

GPQA

(0-shot)

27.46

28.48

31.47

IFEval

65.12

72.23

74.89

IFBENCH

16.41

20.56

20.7

GSM8K

(0-shot)

30.1

46.4

58.3

MGSM

(5-shot)

29.52

45.36

55.04

MMMLU

(5-shot)

37.99

43.28

46.73

*We evaluated LFM2 across seven popular benchmarks covering knowledge (5-shot MMLU, 0-shot GPQA), instruction following (IFEval, IFBench), mathematics (0-shot GSM8K, 5-shot MGSM), and multilingualism (5-shot OpenAI MMMLU, 5-shot MGSM again) with seven languages (Arabic, French, German, Spanish, Japanese, Korean, and Chinese).

Built for efficient inference at the edge.

Unmatched speed, quality, and memory efficiency on the edge.

Whether deploying on smartphones, laptops, or vehicles, LFM2 runs efficiently on CPU, GPU, and NPU hardware. Designed for millisecond latency, offline resilience, and data privacy, LFM2 unlocks the full potential of edge devices across industries.

Phone

Automotive

Drones