All General Coding Reasoning Multimodal Safety Specialized Agentic Language Long Context Speed Text-to-Image Image Editing Text-to-Video Image-to-Video
BFCL83 BrowseComp OSWorld-Verified TAU-bench18 TAU-bench Airline22 TAU-bench Banking Knowledge27 TAU-bench Retail23 TAU-bench Telecom18 Tau Banking (Artificial Analysis)209 Tau2 Telecom (Artificial Analysis)439 Terminal-Bench 2.1241 Terminal-Bench Hard431 ToolBench WebArena

OSWorld-Verified

Agentic

Desktop computer use tasks — GPT-5.4 surpassed human performance (75% vs 72.4%)

Higher is better

No scores recorded yet.