Google has yet another AI model: Gemini 1.5 Flash. It promises to be just as powerful as Pro and Google’s other Gemini models but also way faster.
is a reporter who writes about AI. She also covers the intersection between technology, finance, and the economy.
It should give developers more choices in the models they can use to build their new applications, but since it’s not yet available to consumers, regular Gemini chatbot users can’t use its speedy chat response capabilities yet. Instead, you’ll need to use Google AI Studio to access it.
Gemini 1.5 Flash will work for “narrow, high-frequency, low-latency tasks,” while the Gemini 1.5 Pro model, also soon to be accessible in Google AI Studio, is more suited to activities that don’t depend on quick answers. In human speak, Gemini 1.5 Flash might be a better option for real-time responses to customers or fast image generation, while Gemini 1.5 Pro could read and summarize research papers. Both models are multimodal, meaning they process text, images, and videos.
Originally, there were only three versions: Gemini Pro, the smaller Gemini Nano mainly for devices, and Gemini Ultra, which the company says is the most powerful model it has. Josh Woodward, vice president for Google Labs, told reporters in a briefing before Google I/O that even though Google released its bigger AI model Gemini Ultra in previews, “where we’re really seeing developer interest is in the Pro class of models and this Flash size.”
Gemini 1.5 Flash will now be available for public preview. Both Gemini 1.5 Flash and Gemini 1.5 Pro will have a context window — how much information the model uses at any one time — of up to 1 million tokens (aka words), which is bigger than the 128,000 tokens for GPT-4. A private preview will be available only via a waitlist, with an experimental 2 million context window for Gemini 1.5 Pro.
But Gemini Pro wasn’t totally ignored during the big announcement
Gemini 1.5 Pro is coming to the AI Studio soon, and Google says it updated the model, which is just a few months old, to improve its translation, reasoning, and coding abilities. It will also now be available on Google Workspace, allowing people to use the AI model to summarize emails from Gmail or analyze PDFs. Paid subscribers to Gemini Advanced, the version of Google’s chatbot that uses Gemini Ultra, can access Gemini 1.5 Pro in 35 languages, so they can translate or write out prompts in those languages.
Both models will be available via Google AI Studio and the Gemini API in over 200 countries, including the European Union, the UK, and Switzerland.
Related:
Update, May 15th: The 2 million context window preview will only be available for Gemini Pro at launch, according to Google spokesperson Abby Wolff. Google originally told us it would be offered for Gemini Flash, too.
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
