Skip to content

Dataset · AJ-Gonzalez

Qwen4-Max model reasoning traces on 250 prompts

AJ-Gonzalez/qwen4-max-reasoning-250x

250 complete reasoning chains and final answers from qwen/qwen4-max on the MiMo-2.6-Pro 250x prompt set.

What's inside

The table contains 250 rows in English: one for each prompt from the MiMo-2.6-Pro 250x set with a complete reasoning trace and final answer from the qwen/qwen4-max model. The total generation volume is 1,672,656 output tokens (an average of 6,691 tokens per row, maximum 43,473 tokens).

What you can do with it

  • Fine-tuning smaller models for step-by-step reasoning skills (distillation)
  • Analysis of patterns and reasoning chain lengths in Qwen4-Max class models
  • Qualitative evaluation of LLM reasoning abilities on a complex set of 250 tasks
Tasks
Text generation
Languages
English
Size
up to 1K rows
Downloads this month
351
Likes
58

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("AJ-Gonzalez/qwen4-max-reasoning-250x")