Google's newest AI model tops the charts and costs far less to run

Google has released Gemini 3.1 Pro, an upgrade to its Gemini 3 AI model. It launched in preview on Thursday, meaning an early version that is not the final release. One analysis puts it at the top of a widely watched AI ranking, and says it can cost far less to run than top models from OpenAI and Anthropic.

What changed

On the Artificial Analysis Intelligence Index, Gemini 3.1 Pro sits four points ahead of Claude Opus 4.6. It scores highest on 6 of the 10 tests in that analysis. The biggest gains came in reasoning and knowledge, coding, and hallucination reduction. A hallucination is when an AI states something wrong as if it were true.

It also leads the APEX-Agents leaderboard, which measures how AI handles real professional tasks. Brendan Foody, CEO of Mercor, which built APEX, said the model completes 5 tasks that no model has done before.

Who it affects

Businesses and everyday users who pay for AI tools could benefit if lower running costs reach the price they see. The material does not say what Google will charge, or whether tools you already use will switch to this model.

What is not known yet

• Final pricing for Gemini 3.1 Pro
• When the preview becomes a full release
• How it performs on your own everyday tasks

Rankings are standardized tests. They are useful, but they do not guarantee the best fit for your work. Comparing a few tools on real tasks before committing is still the safer move.