Hands-On Review of Google’s High-Speed AI

GahahgIe Gemini’s Subsequent era household merchandise: Gemini 3.5 is right here!

Gemini 3.5 Flash combines frontier intelligence with real-world motion, supporting quick agent workflows, coding, and multimodal inference whereas sustaining the low latency you’ve got come to anticipate from the Flash collection.

With Gemini 3.5 Professional set to be launched subsequent month, let’s check out the flash mannequin and what it brings.

What’s Gemini 3.5 Flash?

Gemini 3.5 Flash is positioned as a mannequin constructed for actual execution, not simply dialog, with an emphasis on long-term job processing, collaborative subagents, richer UI era, and workflow automation at scale throughout each developer and enterprise environments.

Key options of Gemini 3.5 Flash embody:

Higher efficiency than Gemini 3.1 Professional coding and agent-like job

1M token context window 65k most output token

4x quicker 4 thought ranges when it comes to output tokens/second: minimal, low, medium (new default), excessive Mechanically save ideas all through multi-turn conversations

Methods to entry Gemini 3.5 flash

Gemini 3.5 Flash is at the moment accessible for shopper, developer, and enterprise platforms.

Common customers can entry it by way of the Gemini app or Google Search’s AI mode. Builders can use it by way of Google Antigravity, Google AI Studio’s Gemini API, and Android Studio. Enterprise prospects can entry it by way of Gemini Enterprise Agent Platform and Gemini Enterprise.

This mannequin will not be open supply or weighted, so it’s not accessible by way of Hugging Face, however it’s accessible utilizing the Gemini API. In case you are enthusiastic about operating fashions regionally, you need to use Gemma 4.

Palms-on 1: Prototyping

response:

Copy the code and create the HTML and you’re going to get the next outcome:

Some photographs are lacking and a few buttons additionally do not work. Nonetheless, all this was attributable to lower than 10 seconds!!That makes it much more spectacular. This can be utilized to shortly prototype concepts.

Observe 2: Tough issues

I wish to wash my automobile. A automobile wash is 50 meters away. Is it higher to stroll or drive?

response:

Whereas this will seem to be a no brainer to us, LLMs have lengthy struggled to reply this query accurately.

Palms-on 3: Pace ​​up your visuals

I am fascinated by photographs. Present me a diagram that reveals how compression degrades a picture when it’s transformed to jpeg format a number of occasions.

response:

That is adopted by the following picture exhibiting the discount in high quality.

I used to be having points with picture era within the Gemini app, so I used AI mode as a workaround. It labored effectively and was ready to reply to my query inside 10 minutes.

Be aware: All testing was carried out with a free account within the Gemini app.

remaining verdict

What impressed me most by way of these checks was the pace of response. There have been no responses on this listing that took longer than 10 seconds (the period of time it took for Gemini 3.5 Flash to start out responding).

The standard of the response could be additional improved, however this isn’t an issue for the reason that flash mannequin will not be meant for use for prime quality responses (which take time).

conclusion

Gemini 3.5 Flash appears promising not solely in concept but additionally in outcomes. Versatile in options and pace, the Gemini 3.5 flash mannequin packs numerous punch. It should even be fascinating to see how the Professional model of this mannequin household performs with different fashions with the identical options.

Learn extra: Google’s TurboQuant: Lower mannequin reminiscence utilization in half