<ul><li>Researchers have developed a large language model named s1-32B, which can outperform OpenAI's o1-preview at a lower cost.</li><li>The s1-32B model exhibits test-time scaling and is the most sample-efficient reasoning model.</li><li>The researchers customized Qwen2.5-32B-Instruct using a dataset containing 1,000 prompts and AI-generated answers.</li><li>Using a machine learning method called budget forcing, s1-32B achieved up to 27% higher scores than OpenAI's LLM in math benchmarks.</li></ul>

New LLM developed for under $50 outperforms OpenAI’s o1-preview

Discover more