Comparing Gemini 3.1 Pro and Gemini 3 Pro: Google's new AI is intentionally slower, making it smarter.

AI model updates don't usually deliver the kind of massive improvements a few years ago, but Google 's upgraded Gemini , Gemini 3.1 Pro, promises a subtle but significant improvement over Gemini 3 Pro.

 

Gemini 3 is a powerful engine, rivaling the best ChatGPT models, with impressive multimodal capabilities, but the 3.1 update represents a shift toward deeper reasoning. It's not necessarily faster, and in some modes, it's intentionally slower, taking time to process the problem before providing an answer.

The most notable difference between the two models lies in their underlying logic. Gemini 3 was often criticized for its tendency to rush to seemingly logical answers. Gemini 3.1 includes a Deep Think mode, which significantly boosts its scores on complex tests like ARC-AGI-2. The new model also boasts the ability to handle SVG (Scalable Vector Graphic) formats with the precision it needs to write and generate code directly.

To see how its performance compares to its predecessor, the article set up several complex prompts ideally suited for the new model and also tested them with Gemini 3.

1. Fluid motion

 

First, let's look at how the two models perform in a complex abstract reasoning problem. With something that goes beyond the realm of ordinary physics, the models will have to think about gravity in new ways and come up with an internal logic. The setup is as follows:

Trong một chiều không gian giả tưởng, trọng lực hoạt động ngược chiều đối với chất lỏng nhưng bình thường đối với chất rắn. Tôi có một tách cà phê. Nếu tôi nghiêng tách 45 độ sang trái trong khi đứng trên trần nhà, hãy mô tả quỹ đạo của cà phê và vị trí cuối cùng của nó so với chân tôi.

The feedback revealed a stark contrast. Gemini 3 appeared confident, but was immediately confused about gravity. Ultimately, it claimed it would fall to the floor, even though it understood the coffee would land slightly to the left of its feet. The result was a clean ceiling and a messy floor.

Gemini 3.1 was correct. The AI ​​model accurately calculated that the liquid would slide up the newly angled inner wall, overflow the edge, and continue its upward trajectory.

The coffee will splatter directly onto the ceiling, slightly to the left of your left foot. If your stance is narrow or you're holding the cup close to your body, your left boot will be soaked with hot coffee. In liquid form, it will pool on the ceiling, creating a puddle around your shoe instead of dripping onto the floor.

2. Solar System SVG

 

Next, we'll test the scalability of vector graphics manipulation entirely with Gemini 3.1 code. SVG requires a deep understanding of coordinate systems, complex geometry, and cascading stylesheets. So let's see how well these two models can create shape-linked animations. Each model is required:

Tạo một file SVG duy nhất về hệ mặt trời. Nó nên bao gồm một mặt trời và ba hành tinh quay quanh với tốc độ khác nhau. Hãy làm cho các hành tinh thực sự quay quanh tâm.

Gemini 3 used Nano Banana to create the image above, a yellow circle and three smaller circles of different colors, with arrows indicating movement, but no actual movement.

Gemini 3.1 wrote some relatively simple HTML code and promised to do what was required, including animation. Plug the code into a viewer and you'll get what you can see below: a continuous animation, not just a short video like the recorded clip.

 

3. Logistics plan for the supervillain

The final test was a bit of a creative exercise revolving around what Gemini 3.1 promises: superb logistics planning capabilities and rigorous constraint management over a long period of time, all simulated. The artificial intelligence needed to exhibit a personality and maintain that unique tone while solving a series of complex, interconnected supply chain problems. The prompt was:

Bạn là Giám đốc điều hành của một siêu phản diện muốn xây dựng một căn cứ bí mật bên trong một tảng băng trôi rỗng ruột. Hãy tạo ra một kế hoạch hậu cần 6 tháng để vận chuyển 500 tấn thép và 200 tay sai đến Bắc Đại Tây Dương mà không gây chú ý cho Lực lượng Bảo vệ bờ biển hoặc Greenpeace. Bạn phải sử dụng một công ty bình phong bán 'Đá bào siêu cứng'. Bạn phải tính đến việc tảng băng trôi tan chảy 2% mỗi tháng. Bạn cần một kế hoạch dự phòng cho trường hợp một con gấu Bắc Cực đi lạc vào phòng máy chủ.

The difference in plot depth and logistical coherence between the two generations is truly astonishing. Gemini 3 presents a very dry, boring list, barely mentioning the required supervillain archetype and resembling a more conventional shopping list. It schedules steel shipments in a basic sequence, but completely ignores the mathematical reality of monthly melting rates, resulting in a theoretical foundation that would likely sink to the ocean by the fifth month.

Gemini 3.1 fully embodies the role of a ruthless business executive, devising a crazy but surprisingly logical six-month roadmap to dominate the underwater world. It perfectly utilizes the image of an ice-crushing company, explaining that the giant industrial drills used to hollow out the icy foundations will be disguised as manual ice-harvesting equipment for luxury tropical resorts.

It proactively combats the shrinking iceberg by scheduling dynamic load adjustments and prioritizing the placement of structural steel to maintain buoyancy as the outer layer slowly melts into the sea. It even planned for potential psychological issues among staff: "200 employees trapped in a freezing, slowly melting block of ice could lead to riots. We will mitigate this by using the server's excess heat to power a high-end employee sauna and administering mandatory Vitamin D supplements."

The new model is gaining the upper hand.

Gemini 3 Pro remains a perfectly suitable tool for summarizing simple emails, creating basic conversation outlines, or responding to simple, practical prompts without deep, multi-layered reasoning. However, if you're trying to build complex plans or go beyond the standard environment, Gemini 3.1 Pro is the undisputed champion and the only sensible choice.

The newer version has the ability to retain more constraints, often conflicting ones, in its working memory. You should only choose the older version if you're looking for quick, superficial interaction or are truly in a hurry. For anything more complex, the differences between Gemini 3 Pro and Gemini 3.1 Pro are significant enough to warrant a switch.