nanobench

Navigation

How it works All evals My evals

Legal

Terms of service Privacy policy

Impossible Prompts (version 3)

Comparison
5 prompts
·
13 models
Raters are asked: Select the image that more accurately and completely depicts the prompt: “{prompt}”
Completed
Ratings CSV One row per individual rating Results JSON Eval settings, scores, and every rating Images (zip) Kept for at least 5 years after creation
Clone eval

Add more comparisons

More comparisons tighten the confidence intervals on each model's Elo score.
Total
additional comparisons

Elo scores

Each bar spans a 95% credible interval; the dot marks the mean. Hover for exact values.

Cost vs. Elo score