ModelCap

Current AI model comparison

Exact identities · public evidence

OpenAI

GPT-4o-mini

Position

#139

Index

39.7

Measured

publisher-corpus-prior over 209 held-out anchors (59%); 1 specialist-board observation at 2.1% support (41%); shrunk 25.3 up toward the measured corpus

Comparison side 1

OpenAI

gpt-oss-safeguard-20b

Position

#169

Index

31.3

Inherited

finetune of openai/gpt-oss-20b · fp4 (100%)

Comparison side 2

Output / 1M

$0.60

vs $0.30

Context

128K

vs 131K

Weights

API only

vs Open weights