OrionLLM/GRM-2.6-Air-Opus
OrionLLM/GRM-2.6-Air-Opus is a 1.2 billion parameter language model developed by OrionLLM, fine-tuned from LiquidAI/LFM2.5-1.2B-Thinking. It specializes in Claude Opus-style structured reasoning, making it suitable for clearer and more organized problem-solving. This compact model is optimized for edge-friendly, lightweight local and on-device deployment, offering improved performance on difficult reasoning tasks. It provides practical, lightweight reasoning capabilities within a small footprint.
Loading preview...
GRM-2.6-Air-Opus: Compact Reasoning Model
GRM-2.6-Air-Opus, developed by OrionLLM, is a 1.2 billion parameter model fine-tuned from LiquidAI/LFM2.5-1.2B-Thinking. It is designed to deliver Opus-style structured reasoning in a compact, edge-friendly package, making it ideal for lightweight local and on-device applications.
Key Capabilities
- Opus-Style Reasoning, Compact Size: Distilled from Claude Opus-style reasoning traces, it produces clearer and more organized thought processes despite its small 1.2B parameter count.
- Edge-Ready Deployment: Built on LFM2.5-1.2B-Thinking, this model is well-suited for low-resource environments and on-device use.
- Improved Problem-Solving: Targeted training on harder reasoning traces enhances its performance on complex tasks compared to its base model.
Performance
This model focuses on maximizing structured thinking from a 1.2B model. For instance, it achieves 31.99 on the MMLU benchmark, demonstrating its practical reasoning capabilities in knowledge and STEM domains.