# LFM3-1.7B

liquid-ai, released 2026-10-27

**Verdict: conditional.** LFM3-1.7B is a small open-weight text model from Liquid AI that we would pick for on-device agents that call tools and return structured outputs, or as a base for domain fine-tuning. The deciding caveat: Liquid's own model card says it is not recommended for agentic coding or complex multi-step reasoning. It suits developers shipping edge agents; we have not tested it. Last verified 2026-11-10.

## Pricing

- Input: $0.08 / 1M tokens
- Output: $0.20 / 1M tokens
- Context: 256K ctx; cached input $0.02

## Strengths

- Local agents, tool calling and structured outputs. Liquid's model card lists these as the intended uses.
- Domain-specific fine-tuning of a small model. The model card lists it as an intended use.
- Memory-constrained devices. Liquid says the text model fits in about 1 GB at 4-bit precision.
- Agentic coding or complex multi-step reasoning. The model card says it is not recommended for either "due to its small size".

## Sources

- [Liquid AI LFM3 launch post](https://www.liquid.ai/blog/lfm3)
- [LFM3-1.7B model card (Hugging Face)](https://huggingface.co/LiquidAI/LFM3-1.7B)
- [Liquid AI pricing](https://www.liquid.ai/pricing)

Canonical: https://neomanex.com/models/lfm3-1-7b
