Google has just released the Gemini 3.1 Pro ( Gemini Casino ). This is a deal because it is doing really well on lots of tests that measure how good a language model is. Google made this announcement on February 19 2026. They are saying that Gemini 3.1 Pro is one of the language models out there right now. You can use it now. It is still a preview, which means it is not fully finished yet.
Table of Contents
The new Gemini 3.1 Pro is an update to the Gemini 3 Pro that came out in 2025. That one was already very good. The new one is even better. It is doing a lot better on things like reasoning and solving problems. People who test these models are saying that this is a step forward for Googles AI technology.
When we talk about how good a language model’s we need to look at how it does on real tests. This is because we want to know how it really works, not just what the company says about it. There are lots of tests that measure how good a language model is. Gemini 3.1 Pro is doing very well on many of them.
One of the things that Gemini 3.1 Pro’s really good at is abstract reasoning. This means it can solve problems that it has not seen before. It got a score of 77.1% on a test called ARC-AGI-2. This is more than twice as good as the Gemini 3 Pro. This shows that the new model is really good at thinking and solving problems.
Gemini Casino 3.1 Pro
Gemini 3.1 Pro is also doing better than models on lots of tests. For example it is doing better than Anthropics Claude Opus 4.6 and OpenAIs GPT-5.2 on tests that measure knowledge and science reasoning. It is also doing well on tests that measure how well it can do tasks.
Google has released a lot of data that shows how well Gemini Casino is doing. According to this data it is leading in 13 out of 16 tests. This is a deal because it shows that the model is really good at lots of things.
There are some tests where other models are still doing better. For example on some tests that measure how well a model can reason Claude Opus 4.6 is still doing better. This reminds us that no one model is the best at everything.
People in the intelligence community are paying attention to Gemini 3.1 Pro.
Brendan Foody, the CEO of a company that tests AI models said that Gemini 3.1 Pro is doing well on a test that checks how good a model is at doing professional tasks.
This shows how quickly AI technology is getting better.The Gemini 3.1 Pro model is available now. Only in a preview version.
This means that developers and researchers can start using Gemini 3.1 Pro away.
However, it is not fully completed yet.Gemini Casino will be available to everyone soon.These tests are important because they help us see how good a language model really is.
They give us an idea of what a model, like Gemini 3.1 Pro can do.We can also compare it to models.This helps developers and users choose which model to use for their tasks.
The tests help us understand the strengths of Gemini 3.1 Pro and other models.This way we can pick the model for a specific job.
The release of Gemini 3.1 Pro is a deal because it shows that Google is really committed, to making the best language models. It is also a reminder that the AI community is moving fast and new models are coming out all the time. As these models get better and better we will start to see them being used in real-world applications. That will be really exciting.