Alibaba’s Qwen team has introduced QwQ-32B-Preview, an AI model designed for reasoning tasks with 32.5 billion parameters.

Reasoning tasks refer to those involving answering questions or solving problems across subjects such as mathematics, logic, and visuals.

Available under the Apache 2.0 license, it offers relatively permissive access.

QwQ-32B-Preview can handle prompts up to approximately 32,000 words.

It has outperformed OpenAI’s models, o1-preview, and o1-mini, on benchmarks like Aime and Math.

Aime evaluates AI systems’ performance, while Math tests reasoning and math capabilities.

Recent Alibaba developments