Reaction
GPT-5-based frameworks are claimed to reach 70% on OSWorld
A user compares the claimed score with a 72% human mark and says human-level computer use is within reach.
TLDR
A user says GPT-5-based agentic frameworks reached 70% on OSWorld, a real-computer benchmark for multimodal agents working across operating systems. They compare that score with a 72% human mark and suggest human-level computer use is within reach.
Combined views
33.9K
1 Source, first seen ago
