OppokJr/DeepSeek-R1-Distill-Qwen-1.5B

TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:1.5BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 4, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The DeepSeek-R1-Distill-Qwen-1.5B model by DeepSeek-AI is a 1.5 billion parameter language model with a 32768 token context length, distilled from the larger DeepSeek-R1 model. It is fine-tuned on reasoning data generated by DeepSeek-R1, leveraging a Qwen2.5-Math-1.5B base. This model is specifically designed to transfer advanced reasoning patterns into a smaller, more efficient architecture, making it suitable for applications requiring strong reasoning capabilities in a compact form.

Loading preview...

Overview

DeepSeek-R1-Distill-Qwen-1.5B is a 1.5 billion parameter language model developed by DeepSeek-AI, featuring a 32768 token context length. It is part of the DeepSeek-R1-Distill series, which focuses on transferring the advanced reasoning capabilities of the large DeepSeek-R1 model into smaller, more efficient architectures. This specific model is built upon the Qwen2.5-Math-1.5B base and has been fine-tuned using reasoning data generated by the powerful DeepSeek-R1.

Key Capabilities

  • Reasoning Distillation: Effectively distills complex reasoning patterns from a larger, high-performing model (DeepSeek-R1) into a compact 1.5B parameter model.
  • Enhanced Performance: Demonstrates strong performance on various benchmarks, particularly in mathematical and reasoning tasks, outperforming other models in its size class due to its specialized training.
  • Efficient Deployment: As a smaller model, it offers advantages in terms of inference speed and resource consumption, making it suitable for environments with limited computational resources.
  • Flexible Integration: Can be easily integrated and served using tools like vLLM and SGLang, similar to other Qwen or Llama models.

Good For

  • Resource-Constrained Environments: Ideal for applications where computational resources are limited but strong reasoning is still required.
  • Mathematical and Code Reasoning: Excels in tasks demanding logical deduction and problem-solving, as indicated by its performance on benchmarks like AIME 2024 and MATH-500.
  • Research and Development: Provides a valuable tool for researchers exploring model distillation techniques and the transfer of reasoning capabilities.
  • Applications Requiring Specific Prompts: Recommended usage includes setting temperature between 0.5-0.7, avoiding system prompts, and explicitly instructing step-by-step reasoning for mathematical problems.