decidingbench

Images

How good is it when the state includes a photo or screenshot?

Only for models and providers that accept images. Others show not supported.

Accuracy with images photo quality review

Jev 1.14
TypeSafe
Jev 1.13
TypeSafe
Clef
Cloudflare
Clef-Flash
Cloudflare
pplx-decider v1.2
Perplexity
pplx-decider-flash v1.2
Perplexity
OpenAI Decisions
OpenAI
Microsoft-Decision-1
Microsoft
Kev-4B
Jared Palmer

Accuracy on Wikimedia Commons quality image candidates, for models that accept images.

Test items and precision

Items
500 photo quality reviews and 200 screenshot routing decisions per test set
Precision
About ±4 points on photos; screenshots are indicative only (about ±7).

Full method and scoring rules