Apr 15, 10:58 AM

Google's Gemini 2.5 Pro and Perplexity's Sonar-Reasoning-Pro-High Lead LMArena's Search Arena with 7k Battles Open-Sourced

Google's Gemini 2.5 Pro and Perplexity's Sonar-Reasoning-Pro-High have emerged as top performers in the newly released Search Arena Leaderboard by LMArena, both securing the first position. The leaderboard, which measures the quality of web-search grounded large language model (LLM) completions, showcases the competitive edge of these models in the AI search domain. LMArena has open-sourced 7,000 battles with user votes and provided a detailed analysis in a blog post. Gemini 2.5 Pro, developed by Google DeepMind, has been praised for its strong performance across various benchmarks and internal tests, particularly in long context understanding. Perplexity's Sonar-Reasoning-Pro-High model demonstrated its prowess by outperforming Gemini 2.5 Pro in head-to-head battles 53% of the time, while other Sonar models also ranked highly, dominating the leaderboard. Perplexity is focused on improving its Sonar models and enhancing its search index.

#Google #Search Arena Leaderboard #LMArena #Google DeepMind #Sonar #Perplexity

Written with ChatGPT .