Experimental mini version of OpenAI’s o1 model, optimized for STEM tasks with efficient performance. Not recommended for production use and may be heavily rate-limited.