Google's engineering test ranks Claude first and Gemini fourth

Google ran an engineering test on AI models, and its own Gemini did not win. Claude took first place with a score of 48.3%. Gemini finished fourth, behind three rivals. The results were reported by FourWeekMBA.

What the scores show

The most striking detail is the top score itself. At 48.3%, Claude finished under 50%. That means no AI model aced this exam, and even the leader got more than half of it wrong or incomplete by this measure.

Claude is the AI assistant from Anthropic. Gemini is Google's own AI model. Google's model landing fourth on a test tied to Google is a notable result.

What is not known yet

The source material is thin. It does not say which three models beat Gemini, or what scores they and Gemini received. It also does not explain what tasks the engineering test covered, how it was graded, or when it was run.

• Which models took second and third place
• Gemini's exact score
• What the test asked the models to do
• Whether the results will hold on other tests

What to do

Rankings change from test to test, so one result should not decide your purchase. If you are thinking about paying for an AI tool, try two or three of them on your own work first. Use real tasks from your day, then compare how useful the answers are.