Skip to main content

Google's Gemini 3.8 Flash Set to Launch Wednesday, Taking on OpenAI and Anthropic

Google is about to drop a new weapon in the AI arms race. According to insiders, the tech giant plans to launch Gemini 3.8 Flash as early as Wednesday local time. This isn't just another incremental update—it's a direct challenge to the likes of OpenAI and Anthropic, with a sharpened focus on programming prowess.

The model, which goes by the internal codename "Skimaki," has been turning heads inside Google. In tests conducted with the company's Jet Ski programming tool, engineers showed a clear preference for Skimaki over Anthropic's Opus model. Internal benchmarks suggest that Gemini 3.8 Flash has significantly closed the gap in core coding capabilities, bringing Google within striking distance of its rivals' flagship offerings.

Image

This launch couldn't come at a more pivotal moment. The Google DeepMind team is undergoing a major shake-up. 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 stepped up as Senior Vice President, and he's wasting no time—he's made it clear that the team needs to move faster, much faster.

Google has had its fair share of stumbles lately. Since Gemini 3.0 rolled out last November, the company's models have lagged behind 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 during the summer. That kind of brain drain would rattle any organization, but Google is betting that a fresh approach can turn things around.

So, what exactly is Gemini 3.8 Flash? It's part of the Flash series, which means it's designed to be smaller, faster, and more cost-effective than the heavyweight Pro models. It's not trying to be the trillion-parameter beast that the Pro series represents. In fact, the higher-end Pro model has been delayed for months. The previous 3.5 Pro candidate was scrapped because it didn't offer enough of an improvement over the Flash series. Meanwhile, the next-gen flagship, Gemini 4, has shown promise in pre-training evaluations, but the post-training work is still in progress.

To bolster its AI models from the ground up, Google has been pouring more research time and computing power into reinforcement learning this year. And they've made a key hire: Barret Zoph, co-founder of the former Thinking Machines Lab and ex-head of pre-training at OpenAI, has joined as Research Vice President. He'll be leading the charge on reinforcement learning and post-training breakthroughs.

This move signals that Google isn't just playing catch-up—it's aiming to leapfrog the competition. With a new leader at the helm and a renewed focus on execution, the question is whether Skimaki will be the game-changer Google needs. Only time will tell, but one thing's for sure: the AI battlefield just got a lot more interesting.

Key Points

  • Gemini 3.8 Flash, codenamed Skimaki, is expected to launch as early as Wednesday.
  • The model shows significant improvements in programming, narrowing the gap with Anthropic and OpenAI.
  • DeepMind is undergoing leadership changes, with Koray Kavukcuoglu now at the helm.
  • Google is investing heavily in reinforcement learning and has hired Barret Zoph from OpenAI to lead these efforts.
  • The higher-end Pro model remains delayed, but Gemini 4 shows promise in early evaluations.