didula-wso2/Qwen3-8B-rl_with_think_knowledge_merged

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:32kPublished:May 13, 2026License:apache-2.0Architecture:Transformer Open Weights Warm

The didula-wso2/Qwen3-8B-rl_with_think_knowledge_merged is an 8 billion parameter Qwen3 model developed by didula-wso2, fine-tuned from didula-wso2/Qwen3-8B-ep4_julia_codeforces_extended_with_thinksft_16bit_vllm. This model was trained using Unsloth and Huggingface's TRL library, achieving 2x faster training. It features a 32K context length and is designed for specific applications based on its fine-tuning lineage.

Loading preview...

Model Overview

The didula-wso2/Qwen3-8B-rl_with_think_knowledge_merged is an 8 billion parameter language model developed by didula-wso2. It is a fine-tuned variant of the Qwen3 architecture, specifically building upon the didula-wso2/Qwen3-8B-ep4_julia_codeforces_extended_with_thinksft_16bit_vllm model.

Key Characteristics

  • Architecture: Qwen3-based, with 8 billion parameters.
  • Training Efficiency: This model was trained with a focus on efficiency, utilizing Unsloth and Huggingface's TRL library to achieve 2x faster training speeds.
  • Context Length: Supports a substantial context window of 32,768 tokens.
  • License: Distributed under the Apache-2.0 license.

Potential Use Cases

Given its fine-tuning lineage from a model focused on "julia_codeforces_extended_with_thinksft," this model is likely optimized for:

  • Code Generation and Understanding: Particularly for Julia programming language and competitive programming contexts like Codeforces.
  • Reasoning with "Think Knowledge": The "thinksft" component suggests an emphasis on incorporating reasoning or thought processes, potentially making it suitable for tasks requiring structured problem-solving or logical deduction.

Developers should consider this model for applications where efficient training, a large context window, and specialized capabilities in code-related tasks or reasoning are beneficial.