LFM3-1.7B-Base

Developers and researchers can adapt the checkpoint to specialized domains, languages, or behaviors through continued pre-training or post-training.

Availability

Checkpoint available on Hugging Face

Pricing model

Not established

Deployment

Designed for the edge; 4-bit version takes about 1 GB of memory

Privacy

Not established

Licence

LFM Open License v1.0

• LFM3-1.7B-Base is a pre-trained, text-only base model with 1.73B parameters. ^[1]

  <td style="border: 0; padding: 0.55em 0.75em;">Pre-trained, text-only base model</td>

• The model has a context length of 256K tokens, including input and output. ^[1]

  - **Pre-trained on 22T tokens:** the last 3.5T tokens are curated mid-training data with agentic traces and tool calls, and extend the context window to 256K tokens.

• The checkpoint is intended for continued pre-training or full post-training. ^[1]

  LFM3-1.7B-Base is the starting point for substantial customization, such as adapting to new domains, languages, or behaviors that diverge significantly from those of LFM3-1.7B, through continued pre-training or full post-training via [TRL](https://docs.liquid.ai/lfm/fine-tuning/trl).

• The model card recommends LFM3-1.7B rather than the base checkpoint for chat, tool calling, and agents. ^[1]

  LFM3-1.7B-Base is the **pre-trained checkpoint** behind [LFM3-1.7B](https://huggingface.co/LiquidAI/LFM3-1.7B), a text model for local agents and specialized applications based on our new tri-hybrid architecture.

• The checkpoint is available on Hugging Face under the LFM Open License v1.0. ^[1]

• New models bring live speech translation and voice features Daily brief 2026-10-28