Open-Source AI Model Ranking 2026: Top 20 by Real GitHub Stars

The top 20 open-source AI models ranked by real GitHub stars from the 2026-07-31 snapshot.

## Open-Source AI Model Ranking 2026 Star counts below come from the ainomam GitHub star tracker snapshot dated 2026-07-31. Stars measure ecosystem momentum, not just model quality. ### The top 20 1. **DeepSeek-V3** - 104,067 stars. Frontier-class open weights. 2. **DeepSeek-R1** - 91,984 stars. Reasoning model, OpenAI-o1 rival. 3. **grok-1** - 52,087 stars. xAI open release. 4. **Qwen2.5** - 27,455 stars. Alibaba's flagship open series. 5. **ChatGLM-6B** - 41,012 stars. Historic Chinese open model. 6. **Qwen2.5-Coder** - 16,752 stars. Code-specialized. 7. **DeepSeek-Coder** - 23,991 stars. Code model pioneer. 8. **meta-llama/llama-models** - 7,664 stars. The Llama reference. 9. **mistralai/mistral-inference** - 10,836 stars. Mistral's inference stack. 10. **google/gemma_pytorch** - 5,721 stars. Google's open small models. 11. **internlm/internlm** - 7,255 stars. Shanghai AI Lab. 12. **01-ai/Yi** - 7,822 stars. 01.AI series. 13. **microsoft/Phi-4** - reference repo. Microsoft's small-model line. 14. **NousResearch/Hermes** - community fine-tunes. 15. **togethercomputer/OpenChat** - 8,985 stars. OpenChat 3.5. 16. **Hugging Face leaderboard models** - community entries. 17. **nomic-ai/gpt4all** - 77,402 stars. Local model ecosystem. 18. **lm-sys/FastChat** - 39,507 stars. Vicuna and serving. 19. **oobabooga/text-generation-webui** - 47,507 stars. Local model hub. 20. **ollama library models** - 177,426 stars. The distribution layer. ### How to use this ranking Stars tell you which ecosystem is healthy, not which model is best for your task. Cross-reference with the Hugging Face Open LLM Leaderboard for quality. ### FAQ **Which open model is best right now?** For general tasks DeepSeek-V3 leads open weights; for reasoning DeepSeek-R1; for small local use Qwen3/Qwen2.5. **Are stars a quality signal?** Momentum signal - great for picking an ecosystem, not a benchmark.