OpenAI's GPT-6 Astra Achieves 80% Accuracy in Identifying IKEA Assembly Errors

OpenAI's GPT-6 Astra has demonstrated an 80% accuracy rate in identifying errors in IKEA furniture assembly. This performance starkly contrasts with Claude Opus 4.5, which achieved only 28% accuracy in November 2025, and Claude Fable 5.1 at 70%, while Claude Opus 5 managed 61%. The benchmarking of these models highlights a significant advancement in visual task performance, with GPT-6 Astra completing error identification in just 3 minutes per photo. Manuel Uth noted that "AI can now spot when you've built your IKEA furniture wrong," emphasizing the practical applications of this technology in everyday tasks. Researchers remarked on the notable progress, especially given that earlier models struggled with simpler visual tasks. This development follows the ongoing trend of enhancing AI capabilities in practical scenarios, as seen in previous coverage of AI's evolving role in various applications.