5 prompts that show how the new Gemini 3.5 Flash is its best AI model yet | Tech Radar

Overview

News, deals, reviews, guides and more on the newest computing gadgets

Start exploring exclusive deals, expert advice and more

Details

Unlock and manage exclusive Techradar member rewards.

Unlock instant access to exclusive member features.

Get full access to premium articles, exclusive features and a growing list of member rewards.

5 prompts that show how the new Gemini 3.5 Flash is its best AI model yet

Google promises the latest model can handle creativity and complexity

When you purchase through links on our site, we may earn an affiliate commission. Here’s how it works.

Google showed off a lot of new Gemini features and tools at Google I/O 2026, including the powerful new Google Omni video creator and editor. But the Gemini 3.5 Flash model is supposed to be the real workhorse, according to the company. Google has positioned it as faster and stronger at coding, long context reasoning, multimodal understanding, and more. More importantly, it's supposedly capable of managing the kind of tangled requests real people actually throw at AI.

To see where Gemini 3.5 Flash really stands, I gave it five prompts designed to test very different strengths. Some were practical. Some were deliberately ridiculous. All of them highlighted capabilities Google has emphasized as part of Gemini 3.5 Flash’s evolution beyond earlier Flash models and Gemini 3.1.

For the first test, I wanted to push multimodal reasoning, long context understanding, and technical code generation all at once. Gemini 3.5 Flash is supposed to handle complex information while moving fluidly into practical execution, so I handed it a dense aerospace report about space debris and asked it to make a friendly simulation. Specifically, I wrote, “Use the attached IADC Status of the Space Debris Environment Report to make an interactive simulator showing how debris and orbiting traffic will build up and the potential danger to objects in orbit as a result."

5 prompts for Gemini 3.1 that show off its full potential

Gemini 3.1 Pro vs Gemini 3 Pro: Google’s new AI is slower on purpose

Gemini wrote a long, complicated bit of code using data pulled from the report. When translated, the code became the impressively visual simulation you can see in the video above. It built the interface concept around storytelling rather than raw numbers.

The most impressive part was how clearly it articulated the why behind its choices. “The dashboard should help users understand not simply that debris increases over time, but how launch behavior and mitigation decisions influence long-term outcomes,” Gemini wrote.

I often use trip planning as an AI test because it can show off both the power and flaws in how an AI processes complex variables. Gemini 3.5 Flash emphasizes agentic planning and multi-step reasoning, so the next challenge aimed squarely at seeing how it dealt with a lot of extra details.

I told it to "Plan a 4-day road trip through the Hudson Valley and Catskills. Create a comprehensive, multi-step itinerary that coordinates morning hiking trails, mid-day artisanal food stops, and scenic driving routes, complete with a built-in ‘rainy day backup option’ for each afternoon.”

Gemini 3.5 Flash approached the assignment with surprising restraint. Day one eased into river views and hiking without exhausting the traveler immediately. Scenic routes connected naturally rather than zigzagging unpredictably across the map. Food recommendations aligned geographically quite well, and the weather contingencies made a lot of sense, as Gemini pointed out:

“Rain alternatives should preserve the emotional goal of the original activity. A hiking afternoon replaced by browsing unrelated retail spaces creates disruption rather than continuity.”

Next came procedural reasoning, the kind of structured planning that Gemini 3.5 Flash is supposed to be great at. Thinking about a project I have in mind, I asked Gemini to “act as an expert book conservator and provide a strict, step-by-step amateur guide for case-binding a custom journal at home."

Google reveals dev-focused Gemini 3.1 Flash Lite, promises 'best-in-class intelligence for your highest-volume workloads'

Google I/O 2026 — key announcements from today's massive software showcase

Craft instructions expose weaknesses quickly. Too vague, and beginners fail immediately. Too technical, and people quit halfway through while staring angrily at the glue. Gemini 3.5 Flash found a middle ground, setting expectations and separating essential steps from optional refinements. It accounted for likely mistakes without sounding patronizing.

“Your goal is not museum conservation quality but creating a durable journal while learning foundational binding principles,” it said. “Drying time is part of the process rather than dead time between steps.”

The next test targeted looking at Gemini 3.5 Flash's visual reasoning improvements and claims of better planning for actions. I gave it a picture of a room in my house in need of organizing and cleaning, and told it to "create a 25-minute cleanup plan, tell me what to do first, what to ignore, and how to make the room look 80% better with minimum effort.”

Cleaning advice sounds trivial until you realize most people fail cleanup attempts for strategic reasons rather than motivational ones. Older AI systems often recommend tackling everything equally, which doesn't help matters. Gemini 3.5 Flash understood triage. It said it would prioritize visual impact and momentum.

“Focus first on high-visibility clutter rather than hidden organization problems,” Gemini advised. “Visible progress creates momentum while improving perceived cleanliness rapidly. Avoid opening drawers or beginning deep organization tasks during short cleanup sessions.”

For the final test, I wanted to push Gemini 3.5 Flash’s parallel reasoning, where it breaks a larger problem into smaller pieces and tackles multiple lines of thinking simultaneously rather than solving everything one step at a time.

Just for amusement's sake, I set up a deliberately ridiculous assignment designed to reward structured investigation. I told Gemini to “run a deep, background-agent check on a prospective roommate who claims to be a ‘regular human guy’ but is clearly three penguins stacked inside a trench coat.”

The response leaned into the joke and carried out the mission by splitting the task into parallel investigative tracks and labeling them like a real intelligence operation. One sub-agent handled behavioral analysis. Another focused on environmental evidence. A third examined social consistency signals. Gemini tracked each stream independently while periodically merging findings into an evolving assessment summary.

“Sub-Agent 1: Mobility Analysis: Observed indicators include unusual balance shifts, synchronized lower body movement, and elevated probability of multiple organisms coordinating locomotion.”

Another section read, “Sub-Agent 3: Social Pattern Analysis. Claim of ‘regular human guy’ remains unverified. Additional evidence requested regarding fish purchasing frequency, unexplained ice accumulation, and suspicious resistance toward warm climates.”

Gemini kept the joke going and showed how parallel agentic reasoning changes the shape of AI problem-solving. Earlier systems often handled complicated prompts by thinking through them sequentially, which could make large requests feel slower or less organized. Gemini 3.5 Flash instead approached the fake investigation like multiple specialists collaborating at once.

Gemini 3.5 Flash consistently demonstrated how it could stay oriented on its tasks, something earlier fast models occasionally struggled with. Regardless of whether it was analyzing orbital debris trends, planning road trips, or investigating suspicious penguins, it maintained context while adapting its reasoning style appropriately to the assignment.

The bigger story may be how naturally its strengths meld into a single model. That shift changes what Gemini 3.5 Flash can become in everyday life, at least if people are okay with the trade-offs like needing to give it lots of access to their information to get the most out of it.

Follow Tech Radar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

➡️ Read our full guide to the best business laptops

Best overall: Dell Precision 5690
Best on a budget: Acer Aspire 5
Best Mac Book: Apple Mac Book Pro 14-inch (M4)

Eric Hal Schwartz is a freelance writer for Tech Radar with more than 15 years of experience covering the intersection of the world and technology. For the last five years, he served as head writer for Voicebot.ai and was on the leading edge of reporting on generative AI and large language models. He's since become an expert on the products of generative AI models, such as Open AI’s Chat GPT, Anthropic’s Claude, Google Gemini, and every other synthetic media tool. His experience runs the gamut of media, including print, digital, broadcast, and live events. Now, he's continuing to tell the stories people want and need to hear about the rapidly evolving AI space and its impact on their lives. Eric is based in New York City.

You must confirm your public display name before commenting

1 You've heard of Touch ID and Face ID, but is Ear ID next? Researchers have detailed a new tech would let you use Air Pods or similar buds to prove who you are and unlock your gadgets — and it's actually your heartrate that they detect

2I couldn’t figure out how to delete old Chat GPT images from my Library — here’s the hidden method that finally worked

3'This reveals a broader security problem': Experts warn a key Microsoft legacy tool is still being abused to launch malware campaigns

4 Should you buy a Nintendo Switch 2 before the price hikes? I played 5 of its biggest exclusives to find out

5 Exclusive: The tech keeping Disney Magic Kingdom's most iconic rides running night after night

Tech Radar is part of Future US Inc, an international media group and leading digital publisher. Visit our corporate site.

Key Takeaways

News, deals, reviews, guides and more on the newest computing gadgets
Start exploring exclusive deals, expert advice and more
Unlock and manage exclusive Techradar member rewards
Unlock instant access to exclusive member features
Get full access to premium articles, exclusive features and a growing list of member rewards