Category: Frontier Science Benchmarks

• 30B Model for Scientific Research Outperforms GPT-5.4, Performance Increases from 1.7% to 33.3%