Qwen3.5 Creative Reasoning Benchmark Dataset

A dataset containing text prompts and model completions for evaluating creative reasoning capabilities of the Qwen3.5 large language model. It includes 318,934 total tokens generated by the model, with an average of 2,327.99 tokens per interaction. The dataset was created by Crownelius and was last updated in March 2026.

Use Cases

• Analyze the relationship between prompt characteristics and the length of model completions using the 'Average tokens per row' metric.
• Benchmark the creative reasoning output quality of the Qwen3.5 model against other LLMs using the generated text completions.
• Estimate inference costs for creative text generation tasks using the provided 'Total cost' and token count statistics.
• Study model behavior in single-turn interactions as indicated by the 'Average turns' value of 1.00.

Strengths

• Dataset contains 318,934 total tokens of generated text for analysis.
• Provides specific cost estimation metrics based on OpenRouter pricing.

Limitations

• No column definitions, sample data, or original prompts are provided, limiting reproducibility.
• The dataset's scope is limited to outputs from a single model (Qwen3.5), reducing comparative utility.

Provenance

Source

Crownelius via Hugging Face.

Collection Method

Text generation outputs from the Qwen3.5 large language model.

Freshness

Last updated March 2026.