When it comes to Retrieval-Augmented Generation (RAG) applications, the best options include models like LiquidAI's LFM2, OCC-RAG, and Unsloth-Ornith, each with unique strengths and varying levels of adoption.
Here's a comparison of the top RAG models based on current adoption, maintenance status, and licensing:
| Model | Downloads | Likes | Last Updated | License | Gated | Tags |
|---|---|---|---|---|---|---|
| bartowski/LiquidAI_LFM2-1.2B-RAG-GGUF | 7,223 | 2 | 2025-10-03 | Unknown | No | gguf, conversational |
| LiquidAI/LFM2-1.2B-RAG-GGUF | 3,352 | 61 | 2026-08-05 | Other | No | transformers, text-generation |
| peculiar-ragdoll/Unsloth-Ornith-1.5-35B-A3B | 3,289 | 11 | 2026-08-21 | MIT | No | gguf, conversational |
| occ-ai/OCC-RAG-1.7B | 2,637 | 13 | 2026-06-28 | MIT | No | transformers, RAG |
| bartowski/datagemma-rag-27b-it-GGUF | 1,646 | 2 | 2024-09-12 | Gemma | No | transformers, conversational |
Choose LiquidAI’s models for broad language support and relatively higher adoption, as evidenced by their download counts and positive user interactions. The LiquidAI/LFM2-1.2B-RAG-GGUF model, in particular, has seen 3,352 downloads with the underlying code and architecture tailored for RAG tasks within a transformer library context.
In contrast, the OCC-RAG and Unsloth-Ornith models also merit attention for their unique capabilities (MIT licensed, suitable for various applications), but they haven’t reached the same level of adoption in the community.
Overall, environmental considerations like licensing and active maintenance (e.g., last updated dates) should also inform your choice in the context of your project needs.