Google has unveiled its most powerful Gemini 2.0 Pro AI

Google has launched access to its most powerful AI, Gemini 2.0 Pro experimental model. The AI adds a two million token input window, the largest by any AI from Google, to handle enormous text inputs. Gemini is designed to handle complex prompts on such inputs. Additionally, Gemini 2.0 Pro can search the web and execute code while creating code for apps.

Across a series of standardized AI large language model benchmarks, Gemini 2.0 Pro generally outperforms the company's prior AI models. However, it is still not at the level of humans or the very best AI in all categories tested. For example, on the LiveBench AI LLM benchmark, Gemini 2.0 Pro experimental scores merely 65.13 versus 71.57 for Deepseek R1 or 75.88 for OpenAI o3-mini on high mode.

Nonetheless, for human evaluators comparing AI with their own prompts, Gemini 2.0 Pro ranks among the best two AIs in the world today based on the answers it provided on the OpenLM.ai Chatbot Arena Elo ranking.

Hackers will continue to be disappointed by Gemini 2.0 Pro because the AI was used on itself during training to reduce the possibility of returning non-safe answers.

Gemini 2.0 Pro is available for use by all Google Gemini Advanced users who subscribe at $19.99 per month as well as developers using Google AI Studio and Vertex AI. Readers who want to use Gemini everywhere can install the Gemini app on their phones or purchase a Google Pixel 9 Pro smartphone with Gemini built-in (sold here on Amazon).