Gemini 3.1 Pro: Google's new AI beats ChatGPT and Claude in tests at a lower price

Google has launched Gemini 3.1 Pro, an update to the company's largest AI model. It was released as a "preview", meaning a trial version, on February 19, 2026. The price is the same as its predecessor, Gemini 3 Pro.

Higher score, lower bill

The independent analytics firm Artificial Analysis has measured the models on a combined index consisting of ten different tests. Here, Gemini 3.1 Pro scored 57 points. By comparison, Claude Opus 4.6 at maximum thinking scored 53 points, and OpenAI's GPT-5.2 scored 51 points. Gemini led in six of the ten subtests.

The difference in price is large. According to the analysis, it cost about $892 to run the tests with Gemini 3.1 Pro. For Claude Opus 4.6, the total was $2,486, and for GPT-5.2 it was $2,304. So Google delivers the highest score at a fraction of the cost in this particular test.

What does the model cost?

Companies that use the model in their own systems pay by usage. According to price overviews, Google's price is $2 per million input tokens and $12 per million output tokens. A token is a small piece of text, roughly a short word or part of a long word. For very long texts, the price rises to $4 and $18.

By comparison, Claude Opus 4.7 costs $5 in and $25 out per million tokens, according to a price comparison. The pricing information comes from third-party price sites, so always check Google's own price list.

Not best at everything

The model does not win everywhere. On Arena, a leaderboard for coding, it landed in seventh place. In GDPval-AA, a test of how well AI handles tasks as an agent, it scored 40 percent. Claude Sonnet 4.6 scored 57 percent.

What does it mean for you?

If you use AI at work, a new model does not necessarily change anything right away. But competition pushes prices down, and that may eventually make AI tools cheaper for businesses. Remember, too, that benchmarks are tests and do not always reflect how a model works in daily use.

What it means for you

If your workplace or IT department can choose between several AI models, it may pay to ask them to try Gemini 3.1 Pro, especially if the AI bill is large. If you use AI to write code or solve tasks where the AI works autonomously, it is not the strongest, so test it on your own tasks before switching. It is still a preview, so see it as something you can try out, not as a final choice.