Reasoning benchmark

“Learning objective: find out what reasoning looks like across a spread of models, what reasoning costs and how the models behave on everyday tasks”

The rise of the reasoning model has brought about a massive increase in tokenspend. For most people it was u model upgrade in their favourite chatbot but for companies reasoning and agentic AI has made the costs skyrocket. Most companies pay the bill, but few have a through understanding of what it covers and if they are paying for. Reasoning is hidden to the average user and when the value is unclear it is hard to compare and consider alternatives. That is why I dedided to map the thinking of 13 different models - to show the elephant inside the snake.