Google Introduces Improvements To Gemini 1.5 Variants As 'Newest Experimental Iteration'

The AI field was dull, until Large-Language Models disrupted it.

When OpenAI introduced ChatGPT, the company quickly created an arms race, where other tech companies began competing to either adopt the OpenAI product, or create new ones as competitors. And Google is one of the latter.

After releasing Gemini, the company continues its aggressive updates as it races towards the release of its 2.0 model.

And this time, Google announced a smaller variant of Gemini 1.5, named Gemini 1.5 Flash-8B, along with a "significantly improved" version of Gemini 1.5 Flash and a "stronger" Gemini 1.5 Pro.

According to Google, these models demonstrate increased performance across many internal benchmarks, with "huge gains" in Gemini 1.5 Flash overall and a Gemini 1.5 Pro that excels in math, coding, and handling complex prompts.

"Gemini 1.5 Flash is the best… in the world for developers right now," said Logan Kilpatrick, Product Lead for Google AI Studio, in a post on X.

Read: Google Gemini 1.5 Pro, With Its 'Long-Context Understanding,' Can Understand Details In Films

First of, Google has introduced Gemini 1.5 Flash back in May, which is supposed to be a lightweight version of Gemini 1.5.

The smallest member of the Gemini 1.5 family was designed to handle long contexts and reason over fine-grained information from 10 million tokens and beyond, enabling the models to process high-volume multimodal inputs, including documents, video, and audio.

And this time, Google released an even smaller version of that, and calls it the Gemini 1.5 Flash-8B, trained with 8 billion parameter.

Second, the presence of Flash-8B means that the existing Gemini 1.5 Flash has undergone updates, and has been "significantly improved."

And third, the flagship Gemini 1.5 Pro has been updated to show performance gains in coding and complex prompts and serves as a "drop-in replacement" for its predecessor.

the Large Model Systems Organization (LMSO) posted a leaderboard update to its chatbot arena based on 20,000 community votes. Gemini 1.5 Flash made a “huge leap,” climbing from 23rd to sixth place, matching Llama levels and outperforming Google’s Gemma open models.

Gemini 1.5 Pro also showed “strong gains” in coding and math and “improve[d] significantly.”

Kilpatrick explained in an X thread that these experimental models allow Google to gather feedback and deliver the latest updates to developers as quickly as possible.

"What we learn from experimental launches informs how we release models more widely," he said.

To make matters more streamlined, according to Kilpatrick, Google wants to automatically reroute requests to the new model and will remove the older model from Google AI Studio and the API to "avoid confusion with keeping too many versions live at the same time."

the Large Model Systems Organization (LMSO) that posted a leaderboard update to its chatbot arena based on 20,000 community votes, highly praised the model in a post on X.

"We are excited to see what you think and to hear how this model might unlock even more new multimodal use cases," he posted on X.

"We have been blown away by the excitement for our initial experimental model we released earlier this month," Kilpatrick added. "There has been lots of hard work behind the scenes at Google to bring these models to the world. We can’t wait to see what you build!"

Those who started using it praised Gemini 1.5 Flash for having a "huge leap," and how the 1.5 Pro is showing "strong gains" in coding and math.

Others praised Google for its speedy update and its advancements.

However, a lot of other users questioned why Google had to go to all the troubles with the so many back-to-back updates and not just wait for a 2.0 version.

While the three models show improvements, some critics have called the updates "terrible" and "lazy" for tasks requiring longer outputs, saying that Google is "far behind" Claude, OpenAI, and Anthropic.