LiquidAI/LFM3-1.7B-Base
AI model · https://huggingface.co/LiquidAI/LFM3-1.7B-Base
First seen by Oriel on .
AI summary
Hugging Face hosts LiquidAI/LFM3-1.7B-Base, a 1.73B-parameter text-only checkpoint for heavily customized local assistants and specialized applications, with 256K-token context and edge deployment.
- The model was pretrained on 22T tokens, including 3.5T curated mid-training tokens containing agentic traces and tool calls.
- Its 24-layer tri-hybrid architecture combines 13 gated short convolutions, 5 sliding-window attention layers, and 6 global attention layers.
- At 4-bit precision, it requires about 1 GB of memory, while BF16 inference is supported through Transformers with trust_remote_code=True.
- Use it for continued pre-training or substantial post-training across domains, languages, proprietary data, or novel behaviors; LFM3-1.7B is recommended for chat and agents.
- It supports English, Arabic, Chinese, French, German, Italian, Japanese, Korean, Portuguese, Spanish, Vietnamese, Thai, Indonesian, Hindi, Russian, and Polish under the LFM Open License v1.0.
#language models #edge ai #fine-tuning #transformers
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 140, latest observation ; +32 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 5,266, latest observation ; +1,332 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | LiquidAI/LFM3-1.7B-Base | https://huggingface.co/LiquidAI/LFM3-1.7B-Base Read discussion |