AI Model Drops

When a new AI model comes out, the launch post says it's the best. The posts here check that claim against the numbers.

Every post uses the same chart, so models are easy to compare. Higher on the chart is smarter: that's the Intelligence Index from Artificial Analysis, an independent group that runs every model through the same set of hard tests and publishes the scores.

Further right is cheaper. That's what it cost to run one of those test tasks through the model's API, which is how an app pays for AI: a few cents or dollars each time it asks the model something.

The yellow line is the frontier. A model on the line has nothing above it and to its right, so no other model is both smarter and cheaper. When a new model pushes that line out, it gets a post here and a Short on YouTube.

All numbers come from Artificial Analysis, with the date we pulled them.