Introducing Gemini: Google’s largest and most capable AI model

Google announced Gemini 1.0, which it describes as its most capable and general AI model yet, optimized in three sizes: Ultra, Pro and Nano. Demis Hassabis, CEO and Co-Founder of Google DeepMind, wrote that Gemini was built from the ground up to be multimodal, understanding text, code, audio, image and video. The article states that Gemini Ultra exceeds state-of-the-art results on 30 of 32 widely used academic benchmarks, and that it scored 90.0% on MMLU, making it the first model to outperform human experts on that test. The company also says it trained Gemini on its Cloud TPU v4 and v5e chips and announced the Cloud TPU v5p system.