Written by Wu Jingjing.
Edited by Su Jianxun.
On the evening of February 8, Beijing time, Google finally presented a complete set of products and playing strategies for the large model, and comprehensively faced OpenAI:
renamed the Bard chatbot to Gemini;
release of a dedicated Android Gemini app;
Incorporate all Duet AI features in Google Workspace into the Gemini brand;
The largest and most powerful version of Google's large language model, Gemini Ultra 10 was officially released to the public.
The last time Google made a big mock-up was two months ago with Gemini 10 on the line.
In the future, Google will converge all large model products under the Gemini brand, Bard and Duet will disappear, and Google's large model will have only one name - Gemini. After the united front, Google's large-scale model play is clearer.
gemini
Both Workspace and Duet in Google Cloud will be renamed to Gemini. Duet AI became Gemini for Workspace, and Google One AI Premium users in Workspace can use Gemini in Gmail, Docs, Slides, and Meetings.
The product release on the mobile terminal is the highlight of Google's products, and it also allows ordinary people to directly perceive the functions of the large model through their mobile phones.
Gemini app on mobile.
The newly released Android Gemini app is available directly on Google Play and the App Store** (currently only in the US). On your phone, you can set Gemini as the default assistant, replacing the previous Google Assistant. When you say "Hey Google" to your phone or tablet, you can wake up Gemini, control home devices, set alarms, etc., or you can directly ask Gemini to help write text messages, generate **, and ask questions.
On the Android side, in Google Flights, Google Hotels, Google Maps, Workspace, YouTube and other software, you can also call Gemini assistant operations through "Hey Google".
There is no dedicated Gemini app on iOS, and Google says it will launch a feature in the coming weeks to access Gemini directly from the Google app, with a toggle switch at the top of the app that allows users to chat, draw, and tweet by simply tapping the Gemini switch.
Social** experience of using Gemini.
Mobile products can also help Google win more users, and there are already users who want to be able to help set schedules and create new tasks through direct voice calls.
"I think this is a very important step in building a true AI assistant, a conversational, multi-modal assistant that is more helpful than ever before," said Sissie Hsiao, head of Google Gemini. ”
Currently, Gemini offers both the standard Gemini Pro version and the premium Gemini Ultra version. Gemini Ultra is Google's most powerful AI model at the moment, and has outperformed humans in MMLU (Massive Multitasking Language Understanding) in testing. This model capability will be available in Gemini Advanced and can complete complex tasks such as reasoning, following instructions, coding, and creative collaboration.
Screenshot from Google Gemini
The Gemini Advanced is currently priced at $19$99 per month, the same as GPT-4**, but users can try it for free for 2 months, and Google also gives paid users 2TB of Google Drive storage and all the other features of a Google One subscription. Gemini Advanced is now available in English in more than 150 countries.
One user commented after using Gemini: "It doesn't beat GPT4 because there is no sandbox for running scripts and python file uploads are not allowed." ”
User reviews about Gemini on social media.
In this race for big models, Google is always one step behind compared to its two rivals, Microsoft and OpenAI. Gemini, a multi-modal large model with GPT-4 frontal PK, was only officially launched in December 2023, and Google's large model cloud service was also launched five months later than Microsoft's Azure OpenAI.
Previously, Google has been criticized for its continuous action of large models, but there are too many product types, which is a hodgepodge, and after using Gemini to unite the front, Google also needs to further focus on internal strategy and external development, and coordinate various products in order to win more users.
Right now, Google is making a quick chase. In the previously released Gemini technical report, it can be seen that more than 800 people have participated in technology research and development, relying on a large number of talent reserves, Google is increasing research and development intensity.
After all, next, whether it is GPT4 or GPT5 that has not yet appeared, it is a battle that Google must go all out.