We believe in a world where the default AI path for developers is trustworthy, safe, and open
Learn more
Finetune an LLM using Federated AI
This Blueprint demonstrates how to finetune LLMs using federated AI with Flower, enabling collaboration on private datasets without sharing raw data
Audio Preview
Featured Blueprints
Discover our most popular blueprints.
Text Link
Map Features in OpenStreetMap with Computer Vision
This Blueprint shows you how to fine-tune an object detection model to map features in OpenStreetMap.
Text Link
Fine-tune a speech recognition model for your voice
This Blueprint demonstrates how to fine-tune a Speech-To-Text model in your own language using CommonVoice data and your own voice data.
Text Link
Query structured documents using a lightweight LLM workflow
This Blueprint demonstrates how to use open-source models and a simple LLM workflow to answer questions based on structured documents.
Don’t want to miss these events!
Stay updated on key events shaping the future of open-source development and innovation.
Text Link
Building OS AI with Mozilla.ai - Mar 6th, Stockholm, Sweden
Join us for an event filled with presentations, a three-track mini-hackathon, and networking with fellow AI enthusiasts and experts.
Blueprints Deep Dive - Mar 4th, 6 PM UTC, Discord
Get hands-on with our latest OS AI building blocks.
Scale 22x - Mar 6th-9th, Pasadena, CA
SCaLE is the largest community-run open-source and free software conference in North America.
This week in Open Source
Stay informed with the latest news, trends, and breakthroughs on open-source development.
Text Link
News
Grok 3 Beta — The Age of Reasoning Agents
xAI's most advanced model yet, blending superior reasoning with extensive pretraining knowledge.
Text Link
News
Evo 2, the largest AI model for biology.
A model trained on the DNA of over 100,000 species across the entire tree of life.
Text Link
News
Mistral Saba - a custom-trained model to serve specific geographies
A 24B parameter model trained on curated datasets from across the Middle East and South Asia.
Text Link
News
DeepSeek-AI: Native Sparse Attention
A new mechanism that improves long-context training and inference by reducing costs.
Text Link
News
Nous Research released DeepHermes-3
One of the first models to unify Reasoning and normal LLM response modes into one model.
Text Link
News
Perplexity is open-sourcing R1 1776
R1 1776 is a post-trained version of DeepSeek R1 that doesn’t censor info about China.
Trending in the Hub this week
Explore the open-source resources we use to build our Blueprints.
Tools
Datasets
Common Voice
A multilingual, crowdsourced collection of voice recordings from Mozilla.
Models
OuteTTS-0.1-350M
OuteTTS-0.1-350M is a novel text-to-speech synthesis model.
Qwen2.5-3B-Instruct-GGUF
Qwen2.5-3B-Instruct-GGUF is an instruction-tuned model that generates long-form content, and is optimized for efficient deployment via the GGUF format.
