Gemini 3 Pro Preview High Reasoning 1000X

A dataset of reasoning traces generated using the Gemini 3 Pro Preview model with high reasoning depth. It contains 1000 examples, including some sourced from an existing 250-example dataset, and was created by author REXX-NEW for model distillation purposes. The dataset was last updated on February 25, 2026.

Use Cases

• Fine-tuning open-source LLMs based on the provided summarized reasoning traces.
• Creating distilled versions of the Gemini 3 Pro Preview model.
• Studying the reasoning patterns generated by a high-depth reasoning LLM.
• Benchmarking model performance on tasks requiring multi-step reasoning.

Strengths

• Contains 1000 reasoning examples, providing a substantial base for training.
• Explicitly generated with high reasoning depth, suggesting complex problem-solving traces.
• Includes content from a previously created 250-example dataset, indicating some curation.
• Cost of generation was $32.7 USD for 2.73 million tokens, providing a cost benchmark.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count beyond the stated 1000 examples is unknown, which may limit suitability assessment.
• Data may reflect the biases and limitations inherent to the source Gemini 3 Pro Preview model.

Provenance

Source

Hugging Face, author REXX-NEW.

Collection Method

Synthetically generated using the Gemini 3 Pro Preview API with high reasoning depth.

Time Range

Creation date not specified; last updated February 2026.

Freshness

Last updated 2026-02-25 09:52:35; freshness should be verified.

Geography

Spatial coverage is not applicable for synthetic reasoning data.