---
base_model: unsloth/Qwen3-1.7B
tags:
- text-generation-inference
- transformers
- unsloth
- qwen3
license: apache-2.0
language:
- en
- es
- ja
datasets:
- Stormtrooperaim/Ultra-Thinker-30k
---

# UltraThinker-1.7b 🧠

UltraThinker-1.7b is a fine-tuned version of Qwen/Qwen3-1.7B designed to improve reasoning and output quality. Compared to the base 1.7 billion-parameter model, it’s optimized to “think better” — meaning it produces more coherent, context-aware responses and shows stronger problem-solving and understanding skills in general-purpose tasks.

* Increasing the context window seems to improve the model's performance.



![image](https://cdn-uploads.huggingface.co/production/uploads/684648a88d895eb5ecb537ae/189l86mKZOa9F2sSkRD_-.png)