Mtrini-27B-Tellus

A ~27B QLoRA fine-tune of Qwen3.8-27B for coding, math, reasoning, Arabic and Darija — released as adapter, merged weights, F16 GGUF and IQ2_XS GGUF, with the quantization's importance matrix.