Liquid Nanos

A library of low-latency, task-specific models fine-tuned on Liquid’s multimodal LFM base models. Nanos deliver high accuracy on narrow tasks while remaining small enough to deploy on-device or serve economically at high volume.

Extraction

Structured data extraction from unstructured documents into JSON.

Summarization

Meeting transcripts, document summaries, and content distillation.

RAG & Retrieval

Context-grounded Q&A and fast multi-language document retrieval.

Translation

Low-latency bidirectional translation for short-to-medium text.

LFM2 Nanos

Many Nanos require specific prompting formats to work correctly. See each model’s page for usage guidelines.

LFM2-1.2B-Extract

1.2B · Extraction

Extract structured JSON from unstructured documents.

LFM2-350M-Extract

350M · Extraction

Fastest extraction model for edge deployment.

LFM2-350M-PII-Extract-JP

350M · Extraction

Japanese PII detection into structured JSON.

LFM2-2.6B-Transcript

2.6B · Summarization

Private, on-device meeting summarization from transcripts.

LFM2-1.2B-RAG

1.2B · RAG

Answer questions grounded in provided context documents.

LFM2-ColBERT-350M

350M · Retrieval

Multi-language document embeddings for retrieval and reranking.

LFM2-350M-ENJP-MT

350M · Translation

Near real-time bidirectional Japanese/English translation.

LFM2-350M-Math

350M · Reasoning

Tiny reasoning model for math problem solving.

LFM2-1.2B-Tool

1.2B · Deprecated

Use LFM2.5-1.2B-Instruct for tool calling instead.