(Credit: Tada Images/Shutterstock.com)
Google's free version of Gemini will now use its 1.5 Flash model instead of its 1.0 Pro version, the tech giant announced Thursday.
The 1.5 Flash AI model has a bigger "context window," meaning it can analyze longer documents than before. This also means users can give Gemini a lot more data, and Gemini is expected to respond faster than before while keeping a better "memory" of previous queries.
While this new version is free for all Gemini users, developers still get a more powerful version of 1.5 Flash. Free users get 32K tokens, while invited devs can get 1 million tokens via Google's AI Studio and Vertex AI. In the world of AI, tokens are a unit to measure data or pieces of words.
"The full 1 million token context window is computationally intensive and still requires further optimizations to improve latency, which we’re actively working on as we scale it out," Google said back in February.
But the free, less powerful version of 1.5 Flash can still handle documents like study guides, Google says, noting that it will add the ability to upload files from Google Drive or your computer "soon." At time of writing, Gemini's free web version only allows users to type or paste text, speak via a microphone, or upload photos.
Google's latest Gemini web app update with the 1.5 Flash model is available in 40 languages and over 230 countries and territories (though it's unavailable in China or US-sanctioned nations like Russia or Iran).
Google is also trying to reduce Gemini's hallucinations by getting it to cite other websites when it conveys information as facts. As of Friday, sources cited will appear as gray arrow "chip" icons that can be clicked to display relevant linked content. If you have the Gemini extension in your Gmail, Gemini may also start citing and referencing your email content.
This won't stop hallucinations entirely, however, because not every website on the internet is the ultimate arbiter of truth. Google acknowledges this, adding that its double-check feature may help address contradictions spit out by the AI model.
Those looking for image results may be disappointed, however, as Google still blocks Gemini's ability to create images of people. Back in February, Google said the function would be re-enabled in a "few weeks." It's now been 21 weeks, to be exact.