Gemini 3 Pro Preview
Google · Released 2025-11-18 · 10/17 20:05 updated
See each model's strengths.
Each capability is calculated separately. Insufficient measured evidence is left blank.
Category scores reflect ranking support within their own reference groups. They cannot be added or used to compare absolute levels of different capabilities.
How stable is the overall rank?
Remove one evaluation or organization at a time and vary individual weights and uncertainty handling to see how ranks change.
Insufficient eligibility evidence in 3 scenarios. With the original candidate set fixed, ranks range from 10—51 . This range is not a confidence interval and excludes unpublished results.
Every result has a source.
Public results for this model. Expand an entry to see its run configuration and how it was used.
General evaluations and user experience5 items
Reasoning1 items
Knowledge1 items
Visual understanding1 items
Scored evaluations with no result
These evaluations have not published results for this model. Missing results are not scored as zero.
Some uncertainty remains
Missing evaluations are not scored as zero. Rankings can change with new evidence; avoid overinterpreting small differences between close scores.
Read the methodology →