Trained with SFT (Supervised Fine-Tuning) and PEFT (LoRA) on top of Qwen2.5-1.5B-Instruct, Aether 2.1 provides lightweight and efficient conversational AI performance in German and English.

🔗 Looking for the Base / LoRA Adapter? If you want to use the Hugging Face Transformers PEFT adapter instead, check out the main repository: 👉 Maxilicious20/Aether-2.1

📦 Available Files & Quantizations

Choose the right file depending on your system's VRAM/RAM and performance needs:

Filename	Quantization	Quality	Size	Description / Recommendation
`aether_2_1_fp16.gguf`	FP16 / F16	Maximum	~5.65 MB	Uncompressed full precision adapter GGUF.
`aether_2_1_q8_0.gguf`	Q8_0	Very High	~5.65 MB	High-precision 8-bit quantized build.
`aether_2_1_q4_k_m.gguf`	Q4_K_M	Balanced	~5.65 MB	Recommended. Balanced quantization for ultra-fast local execution.

🚀 How to Run Locally

1. LM Studio

Open LM Studio.
Search for Maxilicious20/Aether-2.1-GGUF or paste the repository ID.
Download your preferred quantization (e.g., aether_2_1_q4_k_m.gguf).
Load the model and start chatting!

2. Ollama / llama.cpp

You can run the GGUF file directly using llama.cpp:

./llama-cli -m aether_2_1_q4_k_m.gguf -p "Hello Aether!" -n 256

Maxilicious20/Aether-2.1-GGUF

Aether 2.1 - GGUF

📦 Available Files & Quantizations

🚀 How to Run Locally

1. LM Studio

2. Ollama / llama.cpp

Maxilicious20/
Aether-2.1-GGUF