🤖 Awesome Open-Source AI Models.

A comprehensive list of open-source AI models... Because nothing says "freedom" quite like code you won't understand anyway!

---

📋 Table of Contents

• Large Language Models (LLMs)
• Image Generation
• Audio / Speech / TTS
• Specialized Models & Tools
• Useful Resources
• Quick Summary
• Contributing

---

🧠 Large Language Models (LLMs)

Alphabetically sorted - because order matters when you're pretending to save the world with AI 🎯

BLOOM

Author: BigScience
License: RAIL License v1.0
Description: The multilingual big sibling, 100% open-source, built by the community. Because unity makes strength... or at least a decent README.

# Example usage
pip install transformers

---

Command R+

Author: Cohere
License: CC-BY-NC 4.0
Description: Even Cohere cracked. Open for commercial use, optimized for RAG and agentic workflows. Turns out capitalists need community love too.

---

Falcon

Author: TII UAE
License: Custom (Apache 2.0 variant)
Description: Falcon-180B was a giant... but the commercial license? A real oriental puzzle 🧩

---

Gemma

Author: Google
License: Gemma Terms of Use
Description: "Open source"... in quotes. More like "open-ish". Free as a fish... in a locked aquarium 🐠🔒

---

Llama 3

Author: Meta
License: Llama 3 Community License
Description: Semi-open license, commercial use allowed... but with terms vaguer than a dream after three Red Bulls.

# Via Ollama
ollama run llama3

---

Mistral 7B / Mixtral 8x7B / Mistral Large

Author: Mistral AI
License: Apache 2.0
Description: French, elegant, and energy-efficient. Mixtral? It's like having 8 brains... but only one pays the bill 🇫🇷💡

Available models:

• Mistral 7B
• Mixtral 8x7B
• Mistral Large

---

OLMo

Author: Allen Institute
License: Apache 2.0
Description: The holy grail: code, data, training logs—everything's public. Transparent as your future after your fifth coffee ☕

---

Phi-3

Author: Microsoft
License: MIT
Description: Small but mighty. Almost open... except you need an Azure account for some weights. So free. So open 🙄

---

Qwen

Author: Alibaba
License: Apache 2.0
Description: China's answer to the rest of the world. Qwen-72B struts in wearing an open-source suit—with a surprisingly permissive license 🐉

---

🖼️ Image Generation

For generating cyborg unicorns in tuxedos... because the world clearly needs that 🦄🤖

Craiyon (formerly DALL·E Mini)

Author: Community
License: Apache 2.0
Description: Open, yes... but the images look like what your cat would draw if it had a keyboard 😹

---

FLUX.1

Author: Black Forest Labs
License: Proprietary (weights available)
Description: Hyper-realistic... but not truly open. Sorry buddy - you're staying outside 🚪

Variants:

• FLUX.1 [pro] - API only
• FLUX.1 [dev] - Non-commercial
• FLUX.1 [schnell] - Apache 2.0

---

Kandinsky

Author: Sber AI
License: Apache 2.0
Description: Russian, open-source, and surprisingly capable. Like a matryoshka doll full of useful features... and vodka 🪆

Available versions:

• Kandinsky 2.2
• Kandinsky 3.0

---

Stable Diffusion

Author: Stability AI
License: Varies by version
Description: The classic. Open weights, commercial use often allowed - but double-check the version before selling your soul 👹

Popular versions:

• SD 1.5 - CreativeML Open RAIL-M
• SDXL - OpenRAIL++-M
• SD3 - Stability AI Community License

# With diffusers
pip install diffusers transformers accelerate

---

🗣️ Audio / Speech / TTS

So your AI can talk in Morgan Freeman's voice... or that of a depressed robot 🎙️

Bark

Author: Suno
License: MIT
Description: Generates speech, music, laughter... sometimes hauntingly. But it's open-source! So it's cool 😅

from bark import generate_audio, SAMPLE_RATE

audio_array = generate_audio("Hello, I am Bark!")

---

Coqui TTS

Author: Community
License: MPL 2.0
Description: Fully open-source TTS framework. Train it on your own voice - if you dare to listen to yourself 🎧

# Installation
pip install TTS

# Usage
tts --text "Hello world!" --model_name tts_models/en/ljspeech/tacotron2-DDC

---

Whisper

Author: OpenAI
License: MIT
Description: The ultimate irony: OpenAI gave us one of the most genuinely open models. Audio-to-text transcription, free and everywhere 🎯

# Installation
pip install openai-whisper

# Usage
whisper audio.mp3 --model medium

---

🧪 Specialized Models & Tools

Because the world needs more than just AIs that say "hello" in 42 languages 🌍

Llama.cpp

Author: Georgi Gerganov
License: MIT
Description: Not a model, but lets you run Llama on a toaster. Almost literally 🍞

# Build
git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp && make

# Run
./main -m ./models/7B/ggml-model.bin -p "Hello world"

---

Nomic Embed

Author: Nomic AI
License: Apache 2.0
Description: Ultra-compact open embeddings. For when you want to shove an AI into a floppy disk 💾

---

Sentence-Transformers

Author: UKP Lab
License: Apache 2.0
Description: Loyal as a golden retriever. Encodes sentences, never judges your life choices 🐕

from sentence_transformers import SentenceTransformer

model = SentenceTransformer('all-MiniLM-L6-v2')
embeddings = model.encode(['This is a sentence', 'This is another one'])

---

🔗 Useful Resources

Ironically, not always that open 😏

Resource	Description	Link
Hugging Face Leaderboard	The official heavyweight championship	→ Visit
Open Source AI Collective	The community that actually shares	→ Visit
Replicate	Test models without selling a kidney	→ Visit
Ollama Library	Your local AI library	→ Visit
Papers with Code	Research papers + implementations	→ Visit

---

📊 Quick Summary

🟢 The Truly Open Ones

Permissive license + code + weights available

• ✅ BLOOM - BigScience (RAIL License)
• ✅ OLMo - Allen Institute (Apache 2.0)
• ✅ Whisper - OpenAI (MIT)
• ✅ Stable Diffusion 1.5 - Stability AI (CreativeML OpenRAIL-M)
• ✅ Mistral 7B - Mistral AI (Apache 2.0)
• ✅ Llama.cpp - Georgi Gerganov (MIT)

🟡 The "It Depends"

Available but with commercial restrictions

• ⚠️ Llama 3 - Meta (Community License)
• ⚠️ Command R+ - Cohere (CC-BY-NC 4.0)
• ⚠️ Qwen - Alibaba (Apache 2.0 but...)
• ⚠️ Mixtral - Mistral AI (Apache 2.0)

🟠 The "Open-ish"

"Open source" with a giant asterisk

• ⛔ Gemma - Google (Proprietary Terms of Use)
• ⛔ Phi-3 - Microsoft (MIT but restricted weights)
• ⛔ FLUX.1 - Black Forest Labs (weights available, restrictive license)
• ⛔ Falcon - TII UAE (complex commercial license)

---

🤝 Contributing

Contributions are welcome! To add a model:

1. Fork this repo
2. Create a branch (git checkout -b feature/new-model)
3. Add your model following the existing format
4. Commit your changes (git commit -am 'Add [Model Name]')
5. Push to the branch (git push origin feature/new-model)
6. Create a Pull Request

Inclusion Criteria

For a model to be added, it must:

• ✅ Have publicly available weights
• ✅ Have clear documentation
• ✅ Be usable without major barriers
• ✅ Have an active community (bonus)

---

📜 License

This repo is under MIT license - Because even sarcastic lists deserve to be free! 🆓

---

⭐ Star History

If this repo was helpful, feel free to give it a star! ⭐

Because a good alphabetical sorting is the foundation of civilization! 🎩✨

---

Last updated: November 2025
Maintained by: The community (and lots of coffee ☕)