Text, audio input-output, vision input-output all need to be at a similar level.

GPT-4o-0513

Text will be measured by ELO on Lmsys with a margin less then 10.

Criterion for audio and vision are yet to be determined.

• Update 2026-03-18 (PST) (AI summary of creator comment): Creator intends to resolve NO unless presented with evidence otherwise, citing that most Chinese models lack audio input/output capability, which is a core requirement.