Google will ask users to pay NZ$36.99 a month for access to its top-of-the line artificial intelligence model, Ultra 1.0.
The pricing is similar to that of Microsoft-backed competitor OpenAI which charges US$20 a month for the Plus version of ChatGPT, with access to the GPT-4 large language model (LLM) and the DALL-E 3 image generator. Microsoft's own-branded Copilot AI which is based on GPT-4, costs NZ$48.50 per user and month, plus GST.
Somewhat confusingly, Google said it has created "a new experience" called Gemini Advanced. This provides access to Ultra 1.0, which you subscribe to as part of the Google One AI Premium Plan.
This is essentially the NZ$16.99 a month Google One Premium Plan for consumers, with two terabytes of storage, and virtual private network access at more than twice the price.
Unlike other Google One plans, AI Premium can't be paid annually so users will forego the 16% discount on offer for the non-AI plans. Gemini is not yet integrated in Google apps like Gmail, Docs, Slides, and Sheets, where it will replace the existing Duet AI.
Users of Google's business-oriented Workspace offering, and Google Cloud, will also have to wait for Gemini to become available for them.
Google has also dropped the Bard moniker for its AI chatbot, which is now called Gemini. The free version of Gemini uses the Pro 1.0 model, which the tech giant says "strikes a good balance between power and efficiency, making it suitable for most general-purpose applications like our conversation today" without requiring excessive resources.
In fact, the free Gemini Pro model suggests that "while there are larger and more powerful versions of Gemini, such as Ultra 1.0, these are typically reserved for very specific use cases due to their computational demands."
Gemini Advanced will also be available on Android and Apple iOS smartphones, via the Google App on the latter operating system, and through Google Assistant or a separate Gemini app.
Google said it has built in guardrails for Gemini, with extensive trust and safety checks, to prevent the AI from providing biased and unsafe responses.