Gemini 3 Pro Preview High Reasoning 250X

484,000 total tokens were generated using the Gemini 3 Pro Preview model with its reasoning depth set to high. The dataset was created by TeichAI for the purpose of fine-tuning open-source LLMs on summarized reasoning traces. It was last updated on January 30, 2026.

Use Cases

• Fine-tuning open-source LLMs based on the provided high-reasoning traces.
• Creating distilled model versions based on the summarized reasoning data.
• Studying instruction-following behavior based on the generated question-answer pairs.
• Benchmarking model reasoning capabilities based on the synthetic dataset.

Strengths

• Created with a high reasoning depth setting on the Gemini 3 Pro Preview model.
• Contains 484,000 total tokens of input and output data.
• Explicitly designed for the concrete task of knowledge distillation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect source bias inherent to the generating model and selected questions.

Provenance

Source

TeichAI via Hugging Face.

Collection Method

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

Time Range

Creation date not specified; last platform update on 2026-01 30.

Freshness

Last updated 2026-01-30 21:20:35; freshness should be verified.

Geography

Spatial coverage is not applicable for this synthetic data.