Google released the AI model ‘ Gemini 3.1 Pro ‘ on February 19, 2026. Google claims that ‘inference capabilities have more than doubled compared to Gemini 3 Pro.’
Gemini 3.1 Pro: Announcing our latest Gemini AI model
https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/
Below is a table listing various benchmark results for ‘Gemini 3.1 Pro’, ‘Gemini 3 Pro’, ‘Claude Sonnet 4.6’, ‘Claude Opus 4.6’, ‘GPT-5.2’, and ‘GPT-5.3-Codex’. Gemini 3.1 Pro recorded the top score in many tests.
Gemini 3.1 Pro achieved a very high score of 98% on
the ARC-AGI-1 test, which measures AI intelligence.
It also achieved a high score of 77% on ARC-AGI-2, demonstrating its high performance at a lower cost than the Claude series and GPT-5 series.
Comparing the scores of ‘Gemini 3.1 Pro’, ‘Gemini 3 Pro’, ‘Claude Sonnet 4.6’, ‘Claude Opus 4.6’, and ‘GPT-5.2’ in ARC-AGI-2 using a bar graph, it looks like this. Gemini 3.1 Pro’s score is more than double that of Gemini 3.Pro.
Artificial Analysis , which analyzes AI performance, also published its analysis of Gemini 3.1 Pro, which showed that in the intelligence category, Gemini 3.1 Pro outperformed competing AIs.
The Gemini 3.1 Pro also has the highest coding performance.
In terms of agent performance, the score was lower than Claude Opus 4.6, GLM-5, and GPT-5.2.
Gemini 3.1 Pro is being gradually rolled out to all Gemini users, and users on the Pro and Ultra plans can also use Gemini 3.1 Pro with NotebookLM.
In addition, Google has published a model card summarizing the Gemini 3.1 Pro’s description and safety information at the following link:
Gemini 3.1 Pro – Model Card — Google DeepMind
https://deepmind.google/models/model-cards/gemini-3-1-pro/