We cracked the code to develop apps powered by local LLMs We created LEAP based on our experience making end-to-end apps for customers. It's model-agnostic, developer-friendly, and available for iOS and Android today. Let's build the future of edge AI
@maximelabonne
-
LFM2-1.2B-Pirate: New Open-Source Language Model
By
–
Notebook: https://
colab.research.google.com/drive/155lr5-u
YsOJmZfO6_QZPjbs8hA_v8S7t?usp=sharing
… LFM2-1.2B-Pirate: https://
huggingface.co/mlabonne/LFM2-
1.2B-Pirate
… -
Fine-tuning LFM2 Models with Axolotl Framework
By
–
You can now fine-tune LFM2 with @axolotl_ai
! Thanks to @winglian for this! You can use his notebook to smoothly SFT an LFM2 model. I tried it and created my own LFM2-1.2B-Pirate based on this code. -
Model Refusal Rate Analysis and Prompt Engineering Support
By
–
Nope, it cannot be training because this behavior is tracked and the refusal rate is the lowest in its size category. I can help more if you send me the prompts.
-
Edge Models Rise: Routers Remain Essential for Distributed AI
By
–
None of this scaffolding will die at the edge.
— Maxime Labonne @ ICLR (@maximelabonne) 14 juillet 2025
Routers will only get increasingly relevant because edge models are catching up.
Keep building smart and local solutions! https://t.co/5p99LhgPJ1None of this scaffolding will die at the edge. Routers will only get increasingly relevant because edge models are catching up. Keep building smart and local solutions!
-
Prompt Engineering Fixes AI Issues Without Retraining
By
–
I can confidently say it's not a training issue. It's hard to tell without your prompt, but that should be easily fixable with prompt engineering. Maybe check again?
-
Hugging Face Hub Conversation Visualization Update
By
–
The conversation viz might be my favorite Hugging Face Hub update ever
-
LFM2 Models Drive App Development and Real-World Usage Growth
By
–
Really cool to see people starting to build apps with LFM2. Benchmarks are important, but real usage is what matters in the end. We were shocked by the popularity of old LFM models on OpenRouter (~400 tks/day). I hope to see many LFM2-powered apps in the next few months!
-
Edge Models Drive Adoption With Apache 2.0 License
By
–
Actually, edge models are (by far) the most demanded ones. The license applies to all derivatives, so it shouldn't be too confusing. We also modeled it after Apache 2.0 for user-friendliness.
-
Balancing Open-Source Contribution with Revenue Generation
By
–
It's possible in the future, but this license was created to balance our desire to contribute to the open-source community with the need to generate revenue. Unlike big tech companies, models are the only way we make money: no ads, no shady practices with user data. I hope