Ishowbackup/DeepSeek-R1-Distill-Qwen-14B

TEXT GENERATIONPricing:Input $0.431 / Cached $0.0862 / Output $1.12Concurrent Unit Cost:1Model Size:14.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 20, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

The Ishowbackup/DeepSeek-R1-Distill-Qwen-14B is a 14.8 billion parameter language model developed by DeepSeek-AI, distilled from the larger DeepSeek-R1 model and based on Qwen2.5-14B. It is specifically fine-tuned using reasoning data generated by DeepSeek-R1 to enhance performance on complex reasoning, math, and code tasks. This model leverages distillation to bring advanced reasoning capabilities to a smaller, more efficient dense model, supporting a context length of 32768 tokens.

Loading preview...

Overview

DeepSeek-R1-Distill-Qwen-14B is a 14.8 billion parameter model developed by DeepSeek-AI, part of a series of distilled models from the larger DeepSeek-R1. DeepSeek-R1 itself is a 671B total parameter (37B activated) Mixture-of-Experts (MoE) model trained with large-scale reinforcement learning (RL) to excel in reasoning tasks, without initial supervised fine-tuning (SFT).

Key Capabilities

  • Enhanced Reasoning: This model inherits and distills the advanced reasoning patterns discovered by the larger DeepSeek-R1, which was trained to perform self-verification, reflection, and generate long chains-of-thought (CoT).
  • Distilled Performance: It demonstrates that reasoning capabilities from larger models can be effectively transferred to smaller, dense models, offering strong performance on benchmarks like AIME 2024, MATH-500, GPQA Diamond, and LiveCodeBench.
  • Qwen2.5 Base: Built upon the Qwen2.5-14B architecture, ensuring compatibility and leveraging its foundational strengths.
  • Long Context: Supports a substantial context length of 32768 tokens, beneficial for complex problem-solving and detailed interactions.

Good for

  • Reasoning-intensive applications: Ideal for tasks requiring logical deduction, problem-solving, and multi-step reasoning.
  • Mathematical and Coding tasks: Shows strong performance in benchmarks related to mathematics (AIME, MATH-500) and code generation/understanding (LiveCodeBench, Codeforces).
  • Resource-efficient deployment: As a distilled model, it offers a more efficient alternative to larger MoE models while retaining significant reasoning prowess.
  • Research and Development: Provides a powerful base for further research into model distillation and reasoning enhancement.

Finetunes
1 models