Model leaderboard

Every model runs the same Glif tests. Scores come from 2,042 blind votes by 10 Glif team members, plus 1,279 AI judge scores, each counting as half a vote. Each kind of task has its own board.

Score out of 100: how often people would pick this model's output over an average model's on the same tests. 50 is average.

How we score

Every model runs the same Glif tests. Team members see two outputs side by side without the model names and pick the one that better fits the test, or call it a tie. An AI judge also scores every output from 0 to 10.

A Bradley-Terry fit turns all those head-to-head results into one rating per model. Ties count as half a win, and each AI judge score counts as half a vote.

The score is the predicted chance people prefer this model's output to the average model's on the same board, so 50 is average. A board is one kind of task, like text-to-image. Scores only compare models on the same board.

The 90% range comes from resampling the votes and refitting many times. Overlapping ranges are close to a tie. A model ranks once it has 20 votes' worth of ratings on a board. Until then it shows no score, only how close it is.

Bars show each score’s 90% range. Overlapping bars are close to a tie.

Image · Brand and text rendering

174 team votes on 6 tests. A model ranks once it has 20 votes’ worth of ratings here.

Score (50 is average)

Credits per run, log scale

Vertical lines show each model’s 90% score range. Horizontal lines show its estimated cost range. Highlighted are the Glif defaults: GPT Image 2.5, Muse Image, Ideogram V4.5.

Not ranked yet (9)