# LFM2.5 1.2B Instruct

> Canonical: https://www.overmindlab.ai/models/lfm-2-5-1-2b-instruct

Sub-2B that rivals larger models for agents, extraction, and RAG.

| Property | Value |
| --- | --- |
| Model ID | LiquidAI/LFM2.5-1.2B-Instruct |
| Provider | Liquid AI |
| Group | LFM2.5 |
| Parameters | 1.2B |
| Tier | Compact |
| Context window | 32K |
| Max training context | 32K |
| Tool calling | Yes |
| Training methods | — |
| Training cost | from $0.50 |
| Serving cost (per 1M output tokens) | from $2.50 |

## Overview

LFM2.5-1.2B-Instruct is Liquid's flagship small instruct model — 28T-token pretrain plus multi-stage RL, ChatML tool calling, and day-one support for Transformers, vLLM, and llama.cpp. Strong on agentic tasks and RAG; not aimed at heavy coding or knowledge-intensive work.

Liquid AI hybrid models for on-device and edge deployment. Fast decode, low memory, and agentic post-training across the ladder.

## Good for

- Agentic tool-calling and data extraction
- RAG on edge or cost-sensitive cloud
- Fast LoRA specialization under 1GB-class footprints

Train this model on your agent's data: https://docs.overmindlab.ai/models/training.md

## Related models

- [LFM2.5 350M](https://www.overmindlab.ai/models/lfm-2-5-350m): 350M, 32K context
- [LFM2.5 230M](https://www.overmindlab.ai/models/lfm-2-5-230m): 230M, 32K context