open-thoughts/OpenThinker3-1.5B
OpenThinker3-1.5B by open-thoughts is a 1.5 billion parameter instruction-tuned causal language model, fine-tuned from Qwen/Qwen2.5-1.5B-Instruct. It features a 131072 token context length and is specifically optimized for advanced reasoning tasks, excelling in mathematical, coding, and scientific problem-solving. This model demonstrates strong performance across various reasoning benchmarks, making it suitable for applications requiring robust analytical capabilities.
Loading preview...
OpenThinker3-1.5B: A Reasoning-Focused Language Model
OpenThinker3-1.5B, developed by open-thoughts, is a 1.5 billion parameter model fine-tuned from Qwen/Qwen2.5-1.5B-Instruct. It is distinguished by its strong performance in reasoning tasks, achieved through training on the comprehensive OpenThoughts3-1.2M dataset. This dataset comprises 850,000 math questions, 250,000 code questions, and 100,000 science questions, with reasoning traces generated by QwQ-32B.
Key Capabilities
- Advanced Reasoning: Demonstrates superior performance in complex problem-solving across mathematics, coding, and science, as evidenced by its benchmark results.
- Benchmark Excellence: Outperforms its base model, Qwen2.5-1.5B-Instruct, and several other 1.5B-class models on benchmarks like AIME24, AIME25, AMC23, and HMMT O2/25.
- Extensive Context: Supports a context length of 131072 tokens, enabling processing of long and complex problem descriptions.
Good For
- Mathematical Problem Solving: Excels in competitive math challenges and general mathematical reasoning.
- Code-Related Reasoning: Strong capabilities in understanding and generating code-related solutions.
- Scientific Inquiry: Suitable for tasks requiring scientific reasoning and knowledge application.
- Research and Development: Ideal for researchers exploring advanced reasoning in smaller language models, particularly given its detailed paper and dataset.