The anticipation for the upcoming release of Google Gemini 2 is reaching a fever pitch. Recent leaks on X have revealed that Google is preparing to launch a new model, Gemini-2.0-Pro-Exp-0111. In a post on X, Google’s senior product manager Logan Kilpatrick hinted at the new model with a subtle nod to OpenAI chief Sam Altman. It is expected to appear under the Advanced section, but it is unclear if it will be for internal testing or a public launch. Users who have tried prompting the model have received fast responses, but it is uncertain if they are truly from version 2.0. This aligns with previous predictions that Google will outperform OpenAI’s o1 model. A user on X reported that an unknown Gemini model was available in the LMSYS Arena and outperformed their test with OpenAI o1-mini. Meanwhile, AI insider Jimmy Apples shared a scoop that Gemini 2.0 may have already been deployed to select B2B customers. Similar to Gemini 1.5, Gemini 2 will continue to generate images and perform web searches, features that will help Google compete with OpenAI’s Search GPT and Perplexity AI. It is also expected that Meta will join the search race. Recently, Google AI Studio and the Gemini API introduced ‘Grounding with Google Search’, allowing developers to improve response accuracy by incorporating real-time data from Google Search. This feature will likely be included in Gemini 2 and its APIs. A user on X who attended Kilpatrick’s session in San Francisco revealed that Gemini 2 will be a larger model with multi-turn capabilities, vision, audio, embeddings, and more. Taking inspiration from Anthropic, Google plans to release a new feature called ‘Jarvis’ that can take control of a user’s web browser to perform tasks like research, purchasing, and booking. This feature will also be integrated into Gemini 2. Similarly, Microsoft is testing Copilot Vision, a feature that enables it to act as an expert tech journalist.
