Fine-tuning Liquid AI LFM2/LFM2.5 MoE models with the new Halo framework
Melvin Vivas · X post · 2026-09-25 · Open on X
Topics: Fine-tuning & Model Customization · Level: advanced
Summary
Melvin Vivas shares Halo, a new fine-tuning framework, along with Liquid AI's two official recipes for fine-tuning their Mixture-of-Experts models. One recipe is an SFT notebook for LFM2.5-8B-A1B in the Liquid4All cookbook. The other is a Halo cookbook for LFM2-24B-A2B.
Key points
- Halo is a newly released fine-tuning framework.
- Liquid AI published recipes for fine-tuning two of its MoE models with Halo: LFM2.5-8B-A1B and LFM2-24B-A2B.
- The names show total vs. active parameters: 8B total with about 1B active, and 24B total with about 2B active.
- The LFM2.5-8B-A1B recipe is a supervised fine-tuning (SFT) notebook in Liquid4All's cookbook repo.
- The LFM2-24B-A2B recipe is a markdown cookbook in Halo's own docs (human-docs/cookbooks).
Resources mentioned
- Halo · tool · github.com · check price
A new fine-tuning framework that has recipes for MoE models like Liquid AI's LFM2 family. - SFT MoE with Halo notebook (LFM2.5-8B-A1B) - Liquid4All cookbook · repo · github.com · free
Notebook showing supervised fine-tuning of the LFM2.5-8B-A1B MoE model with Halo. - Halo LFM2 MoE cookbook (LFM2-24B-A2B) · docs · github.com · free
Halo cookbook explaining how to fine-tune the LFM2-24B-A2B MoE model. - LFM2.5-8B-A1B · tool · liquid.ai · free
Liquid AI's Mixture-of-Experts model with 8B total and about 1B active parameters. - LFM2-24B-A2B · tool · huggingface.co · free
Liquid AI's Mixture-of-Experts model with 24B total and about 2B active parameters. - Liquid AI (@liquidai) on X · website · x.com · free
Liquid AI's family of efficient foundation models, used here as the example of a model you can run in the browser with ONNX.
Also in: Trying Liquid AI Decision Models (d1, d1 ultra) via API (Melvin Vivas on X · notes), Liquid RIPL: Low-Latency Inference Engine for LFMs (Melvin Vivas on X · notes), ONNX + Transformers.js: Run Models Like Liquid LFM in the Browser (Melvin Vivas on X · notes), Try Liquid AI's open d1-3B decision model in Colab (Melvin Vivas on X · notes) and 30 more
Try this
- Run the SFT MoE with Halo notebook to fine-tune LFM2.5-8B-A1B.
- Follow the Halo cookbook to fine-tune LFM2-24B-A2B.
More in Fine-tuning & Model Customization
- Run Local Models on a Free GPU with Google Colab (T4)
- Free Colab Notebooks for Local Model Fine-Tuning and Inference
- Train Your Own LLM Model Router by Fine-Tuning ModernBERT as a Classifier
- Hugging Face LLM Course: Chapter 7.3 for Training Your Own Model
- Model Routing: Fine-Tune Your Own Router on Your Prompts
- Base vs fine-tuned Gemma 4 E2B as a model router