Qwen3-1.7B-Stratos-Thinking (GGUF)
A reasoning-enhanced fine-tuned version of Qwen/Qwen3-1.7B, trained on DeepSeek-R1 distilled thought chains from the Bespoke-Stratos-17k dataset.
Model Details
- Base Model:
Qwen/Qwen3-1.7B - Fine-tuning Method: QLoRA (
r=16,alpha=16) via Unsloth - Training Dataset:
bespokelabs/Bespoke-Stratos-17k(1,500 high-quality reasoning samples) - Quantization Format:
Q4_K_M(GGUF) - Supported Languages: 29+ natural languages (inherited from Qwen3)
- Reasoning Format: Outputs step-by-step internal thoughts (
<think>/<|begin_of_thought|>) before generating the final solution.