Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html> <h"... is not valid JSON

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

SuperiorThoughts-1 Qwen 3.8 Max Reasoning Distill

SuperiorThoughts

10,000 filtered and deduplicated reasoning-heavy traces generated with maximum reasoning effort from Qwen 3.8 2.4T A95B.

Average Trace Size
2 KB
Dataset Size
20 MB
Total Rows
9640
Domains
Math, Code, and General Reasoning.
Executive Summary

SuperiorThoughts-1 is a large-scale synthetic dataset designed specifically for deep reasoning SFT and LLM distillation. Generated using Qwen 3.8 2.4T A95B operating at maximum reasoning capacity, each sample contains a different reasoning trace across multiple domains.

Intended Use Cases

  • Reasoning Distillation: Distilling any LLM in general on Qwen 3.8 2.4T A95B reasoning traces.
  • SLM Distillation: Inject high-quality reasoning data into 1B to 3B parameter models during SFT or continued pre-training.

Downloads last month
70