trinityomnis/OpenThinker-32B
TEXT GENERATIONPricing:Input $2.72 / Output $4.8Concurrent Unit Cost:2Model Size:32.8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 7, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold
OpenThinker-32B is a 32.8 billion parameter language model developed by trinityomnis, fine-tuned from Qwen2.5-32B-Instruct. It specializes in advanced reasoning tasks, demonstrating strong performance on benchmarks like AIME, MATH, and GPQA Diamond. This model is optimized for complex problem-solving and analytical applications, leveraging the OpenThoughts-114k dataset for its enhanced capabilities.
Loading preview...
OpenThinker-32B: Enhanced Reasoning Model
OpenThinker-32B is a 32.8 billion parameter language model, fine-tuned from Qwen2.5-32B-Instruct. It was developed by trinityomnis using the proprietary OpenThoughts-114k dataset, which is derived by distilling DeepSeek-R1. This model is specifically designed to excel in complex reasoning and mathematical tasks.
Key Capabilities & Performance
- Advanced Reasoning: Demonstrates strong performance across challenging reasoning benchmarks, including AIME24 I/II (66.0), AIME25 I (53.3), MATH500 (90.6), and GPQA Diamond (61.6).
- Open-Source Ecosystem: The project emphasizes full transparency, with publicly available model weights, datasets, data generation code, evaluation code (Evalchemy), and training code (LLaMA-Factory).
- Efficient Fine-tuning: Trained for 3 epochs on the OpenThoughts-114k dataset with a 16k context length, utilizing AWS SageMaker and the Leonardo Supercomputer for efficient scaling.
Why Choose OpenThinker-32B?
- Superior Reasoning Benchmarks: Outperforms several 32B models on key reasoning metrics like MATH500 and GPQA Diamond, making it suitable for applications requiring high-level analytical thought.
- Transparent Development: Its fully open-source nature allows developers to inspect, reproduce, and build upon its components, fostering community collaboration and trust.
- Specialized for Complex Problems: Ideal for use cases demanding robust mathematical problem-solving, logical deduction, and general reasoning capabilities.