According to @GoogleDeepMind, the new Gemini 2.5 Pro preview has achieved a +24 LMArena Elo score over its predecessor, showing significant advancements in AI performance. The model leads in challenging coding benchmarks such as AIME and AIDER, as well as in science (GPQA) and reasoning (HLE) evaluations. Improvements in style and structure are attributed to user feedback, reflecting a focus on practical AI applications for developers and businesses. These upgrades position Gemini 2.5 Pro as a competitive solution for enterprises seeking state-of-the-art AI for complex technical and scientific tasks (source: goo.gle/4kKynYo).