schwyzquants/OpenThinker-32B
OpenThinker-32B is a 32.8 billion parameter language model developed by schwyzquants, fine-tuned from Qwen2.5-32B-Instruct. It is specifically trained on the OpenThoughts-114k dataset, which is distilled from DeepSeek-R1, to enhance its reasoning capabilities. The model demonstrates strong performance in mathematical reasoning and general question answering, achieving 90.6% on MATH500 and 61.6% on GPQA Diamond. With a context length of 32768 tokens, it is suitable for complex analytical tasks requiring deep understanding and logical inference.
Loading preview...
OpenThinker-32B: Enhanced Reasoning Model
OpenThinker-32B is a 32.8 billion parameter language model developed by schwyzquants, built upon the Qwen2.5-32B-Instruct architecture. Its core differentiator is its fine-tuning on the proprietary OpenThoughts-114k dataset, which is derived from distilling DeepSeek-R1. This specialized training aims to significantly boost the model's reasoning and problem-solving abilities.
Key Capabilities and Performance
- Superior Reasoning: Achieves notable scores on reasoning benchmarks, including 90.6% on MATH500 and 61.6% on GPQA Diamond, outperforming several comparable models.
- Open-Source Ecosystem: The project emphasizes transparency, providing open access to its model weights, datasets, data generation code, evaluation tools (Evalchemy), and training configurations.
- Robust Training: Fine-tuned for 3 epochs with a 16k context length using LlamaFactory, leveraging substantial computational resources (AWS SageMaker with 8xH100 P5 nodes).
What Makes It Different?
Unlike many general-purpose LLMs, OpenThinker-32B's unique training on the OpenThoughts-114k dataset, specifically designed for reasoning, positions it as a strong contender for tasks requiring logical inference and complex problem-solving. Its fully open-source nature also provides developers with complete transparency and control over its components.
Should You Use This Model?
OpenThinker-32B is particularly well-suited for applications demanding high-accuracy reasoning, such as:
- Mathematical problem-solving and theorem proving.
- Complex question answering and knowledge inference.
- Scientific research assistance and logical deduction tasks.
Developers seeking a powerful, transparent, and specialized model for reasoning-intensive workloads will find OpenThinker-32B a valuable tool.