densafaradenza/DeepSeek-R1-Distill-Qwen-32B

TEXT GENERATIONPricing:Input $2.72 / Output $4.8Concurrent Unit Cost:2Model Size:32.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 12, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

DeepSeek-R1-Distill-Qwen-32B is a 32.8 billion parameter language model developed by DeepSeek-AI, distilled from the larger DeepSeek-R1 model and based on the Qwen2.5 architecture. It is specifically fine-tuned using reasoning data generated by DeepSeek-R1, aiming to transfer advanced reasoning patterns to a smaller, dense model. This model excels in complex reasoning tasks across math, code, and general knowledge, outperforming OpenAI-o1-mini on various benchmarks.

Loading preview...

DeepSeek-R1-Distill-Qwen-32B Overview

DeepSeek-R1-Distill-Qwen-32B is a 32.8 billion parameter model from DeepSeek-AI, part of their DeepSeek-R1 series. This model is a distilled version of the larger DeepSeek-R1, which itself was developed using a novel large-scale reinforcement learning (RL) approach to enhance reasoning capabilities. The core idea behind this distilled model is to transfer the sophisticated reasoning patterns learned by the powerful DeepSeek-R1 into a more compact, dense model based on the Qwen2.5 architecture.

Key Capabilities and Differentiators

  • Reasoning Distillation: It leverages reasoning data generated by the advanced DeepSeek-R1 to imbue smaller models with superior reasoning abilities, demonstrating that complex reasoning can be effectively transferred.
  • Strong Performance: Benchmarks show that DeepSeek-R1-Distill-Qwen-32B achieves competitive results, notably outperforming OpenAI-o1-mini across various math (AIME 2024 pass@1: 72.6, MATH-500 pass@1: 94.3), code (LiveCodeBench pass@1: 57.2), and general reasoning tasks (GPQA Diamond pass@1: 62.1).
  • Optimized for Reasoning: The model is specifically designed to handle complex problems requiring chain-of-thought reasoning, benefiting from the RL-driven exploration of reasoning paths in its parent model.
  • Qwen2.5 Base: Built upon the Qwen2.5-32B base model, it inherits its robust language understanding and generation capabilities, further enhanced by specialized reasoning fine-tuning.

Usage Recommendations

To achieve optimal performance, users are advised to:

  • Set temperature between 0.5-0.7 (0.6 recommended).
  • Avoid system prompts; include all instructions in the user prompt.
  • For math problems, include a directive like "Please reason step by step, and put your final answer within \boxed{}".
  • Enforce the model to start its response with "\n" to ensure thorough reasoning.

Finetunes
1 models