Arc-Intelligence/ATLAS-8B-Instruct
Full fine-tune vs Qwen/Qwen3-8B. Fine-tune: 92% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 0/1, Norms 125/145, MLP 108/108, Attention 144/144, lm_head 1/1
4 derivatives checked
- mradermacher/ATLAS-8B-Instruct-i1-GGUF: Quantization
- mradermacher/ATLAS-Teach-8B-Instruct-i1-GGUF: Quantization
- mradermacher/ATLAS-Teach-8B-Instruct-GGUF: Same weights, converted
- mradermacher/ATLAS-8B-Instruct-GGUF: Same weights, converted