8 13 hours ago

Fine-tuned derivative of Qwen3.5-4B, trained on GPT 5.6 Sol, Luna, and Fable 5.1 traces