New AI-startup Trillion Labs launches ‘rBridge’ method to predict LLM performance

South-Korea based AI startup Trillion Labs today announced the launch of ‘rBridge’, a novel benchmarking methodology that uses proxy models (under 1 billion parameters) to predict reasoning performance of large language models (LLMs). The company says rBridge will allow enterprise users to benchmark and compare LLM reasoning speed and accuracy without the cost-intensive compute loads of massive models. The release comes amid a broader shift toward cost efficiency in enterprise AI and may drive adoption of mid-sized models for production workloads.