QwQ-32B is a 32.5 billion parameter causal language model from Qwen, part of the Qwen series, specifically designed for enhanced reasoning capabilities. It utilizes a transformer architecture with RoPE, SwiGLU, RMSNorm, and Attention QKV bias, and supports a full context length of 131,072 tokens with YaRN for long inputs. This model is optimized for complex problem-solving and achieves competitive performance against other state-of-the-art reasoning models.
Loading preview...
Overview
QwQ-32B is a 32.5 billion parameter causal language model developed by Qwen, focusing on advanced reasoning. It is built on a transformer architecture incorporating RoPE, SwiGLU, RMSNorm, and Attention QKV bias. The model undergoes both pretraining and post-training, including Supervised Finetuning and Reinforcement Learning, to significantly enhance its performance on challenging downstream tasks.
Key Capabilities
- Enhanced Reasoning: Specifically designed to excel in reasoning tasks, offering competitive performance against models like DeepSeek-R1 and o1-mini.
- Extended Context Window: Supports a full context length of 131,072 tokens, with YaRN enabling effective handling of inputs exceeding 8,192 tokens.
- Optimized for Problem Solving: Capable of thoughtful output generation, with recommended guidelines for sampling parameters (Temperature=0.6, TopP=0.95, TopK 20-40) and output standardization for math and multiple-choice questions.
Good For
- Complex Reasoning Tasks: Ideal for applications requiring deep logical inference and problem-solving.
- Long Context Processing: Suitable for scenarios where processing extensive documents or conversations is necessary, leveraging its 131,072-token context window.
- Benchmarking and Research: Provides a robust foundation for evaluating and developing advanced reasoning AI systems.
Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.