Skip to main content

Google's Gemini 3.8 Flash: A New Challenger in the AI Arena

Google is about to make a splash in the AI world with the imminent release of Gemini 3.8 Flash, a model that's been turning heads internally. According to insiders, the tech giant plans to roll out this new model as early as Wednesday, and it's already showing promise in programming tasks.

Codenamed "Skimaki," Gemini 3.8 Flash has been impressing engineers in internal tests, particularly when compared to Anthropic's Opus model. The buzz is that it significantly narrows the gap between Google's offerings and the flagship models from Anthropic and OpenAI, at least when it comes to coding.

Image

But this launch isn't just about a new model—it's happening against a backdrop of significant changes at Google DeepMind. Last month, Demis Hassabis, co-founder of DeepMind, stepped back from day-to-day management to take on the roles of department chairman and Alphabet's Chief Scientist. His long-time deputy, Koray Kavukcuoglu, has taken the reins as Senior Vice President, and he's already made it clear that the team needs to pick up the pace.

Google has had its fair share of setbacks in the AI race. Since Gemini 3.0 dropped last November, the company's models have lagged behind competitors in key areas like programming. To make matters worse, several core researchers, including Noam Shazeer (co-founder of Character.AI) and Chief Scientist Jeff Dean, left the company over the summer.

So, what's the deal with Gemini 3.8 Flash? It's part of the Flash series, which means it's smaller, faster, and cheaper than the Pro models. It's not meant to compete with the trillion-parameter behemoths, but it's a strategic move to stay relevant in the fast-paced AI landscape.

The higher-end Pro series, on the other hand, has hit some snags. The release of the next Pro model has been delayed for months, and the previous 3.5 Pro candidate was scrapped because it didn't offer enough improvement over the Flash series. Meanwhile, the next-gen flagship, Gemini 4, has shown promise in pre-training tests, but it's not ready for prime time yet.

To bolster its AI capabilities, Google has been pouring more resources into reinforcement learning this year. And in a notable hire, the company brought on Barret Zoph, co-founder of the former Thinking Machines Lab and ex-head of pre-training at OpenAI, as Research Vice President. He's now tasked with driving breakthroughs in reinforcement learning and post-training.

So, will Gemini 3.8 Flash be the game-changer Google needs? It's too early to tell, but the company is clearly pulling out all the stops to stay in the race. With a new leader at the helm and a renewed focus on execution, Google is signaling that it's not backing down from the fight.

Key Points

  • Google is launching Gemini 3.8 Flash, codenamed "Skimaki," possibly as early as Wednesday.
  • The model shows improved programming capabilities, narrowing the gap with rivals.
  • Leadership changes at DeepMind, with Koray Kavukcuoglu taking over and demanding faster execution.
  • Google has faced setbacks, including researcher departures and delays in the Pro series.
  • The company is investing in reinforcement learning and has hired Barret Zoph to lead these efforts.