Six open-source models, combined in layers, just beat GPT-4o. AlpacaEval 2.0 score: 65.1% — outperforming GPT-4o’s 57.5% by a 7.6-point margin. That’s what Together AI’s Mixture […]