Gemini 3 Pro Preview High Reasoning 250X

sarayoot's reasoning dataset was created using the Gemini 3 Pro Preview API with a high reasoning depth setting. The dataset contains 484,000 total tokens generated for a cost of $5.78 and was last updated on February 13, 2026. It is intended for fine-tuning open-source large language models on summarized reasoning traces.

Use Cases

• Fine-tuning smaller models for improved reasoning based on high-depth reasoning traces.
• Creating distilled versions of proprietary LLMs based on the provided reasoning summaries.
• Studying the structure and patterns of reasoning generated by advanced LLMs.
• Benchmarking model performance on complex reasoning tasks derived from the dataset.

Strengths

• Dataset creation cost and token volume are explicitly quantified at $5.78 and 484,000 tokens.
• Source model and reasoning depth are specified as Gemini 3 Pro Preview with a 'high' setting.
• Dataset has a clear, stated purpose for knowledge distillation and instruction-tuning.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific question sources are unknown.

Provenance

Source

sarayoot via Hugging Face.

Collection Method

Generated using the Gemini 3 Pro Preview API, with some questions sourced from reedmayhew.

Time Range

Creation date is unspecified; last platform update was February 13, 2026.

Freshness

Last updated 2026-02-13 07:36:23; freshness should be verified.

Geography

Spatial coverage is unspecified.