Openintelligent123/DeepSeek-R1-Distill-Qwen-32B
Openintelligent123/DeepSeek-R1-Distill-Qwen-32B is a 32.8 billion parameter distilled language model from DeepSeek-AI, based on the Qwen2.5 architecture with a 32768 token context length. It is fine-tuned using reasoning data generated by the larger DeepSeek-R1 model, demonstrating exceptional performance on mathematical, coding, and general reasoning benchmarks. This model is optimized for complex problem-solving and outperforms OpenAI-o1-mini across various tasks, making it suitable for applications requiring strong analytical capabilities.
Loading preview...
DeepSeek-R1-Distill-Qwen-32B: Reasoning-Enhanced Distilled Model
This model is a 32.8 billion parameter variant from the DeepSeek-R1-Distill series, developed by DeepSeek-AI. It is a distilled version of the powerful DeepSeek-R1 reasoning model, fine-tuned on the Qwen2.5-32B base using reasoning patterns generated by the larger DeepSeek-R1. This approach allows smaller models to inherit advanced reasoning capabilities without direct large-scale reinforcement learning.
Key Capabilities
- Enhanced Reasoning: Achieves strong performance across mathematical, coding, and general reasoning tasks, outperforming models like OpenAI-o1-mini.
- Distilled Intelligence: Leverages reasoning data from the 671B parameter DeepSeek-R1, demonstrating that complex reasoning can be effectively transferred to smaller, dense models.
- Optimized for Performance: Shows competitive results on benchmarks such as AIME 2024 (72.6% pass@1), MATH-500 (94.3% pass@1), and LiveCodeBench (57.2% pass@1).
- Long Context: Supports a context length of 32768 tokens.
Usage Recommendations
- Prompting: Avoid system prompts; include all instructions within the user prompt. For math, use "Please reason step by step, and put your final answer within \boxed{}".
- Temperature: Recommended temperature range is 0.5-0.7 (0.6 is ideal).
- Enforce Thinking: To ensure thorough reasoning, enforce the model to start its response with "\n".