GPT-6 Astra: visual diagnostics leap to 80% accuracy

Alex da Cruz
Alex da Cruz is a full-stack developer based in São Paulo, Brazil. He works with React, TypeScript and automation, and uses AI daily to solve real problems in code and operations — not as a demo. He has run an e-commerce operation end to end, and now builds and maintains the automation pipeline behind this blog. He writes about what he actually tests.
OpenAI's GPT-6 Astra marks a sharp shift in how multimodal models handle complex physical instructions. According to Epoch AI's Furniture Assembly Benchmark, the model inspects photos of furniture builds containing deliberate errors and matches them against manuals.
How fast is GPT-6 Astra at error detection?
The system hits an 80 percent accuracy rate in spotting mistakes, a major leap from the 28 percent ceiling seen by top models like Claude Opus 4.5 in November 2025. Competitors trail closely in the latest tests, with Claude Fable 5.1 at 70 percent and Claude Opus 5 at 61 percent, while open-weight models like Kimi K3 remain roughly seven months behind.
What is the practical bottleneck for work tasks?
Despite the high accuracy, processing takes about three minutes per photo according to Epoch AI. This processing time prevents real-time assistance on the factory floor or during live repairs, though it points toward reliable offline diagnostics for appliance fixes and technical support soon.
Sources
Frequently asked questions
- What is GPT-6 Astra's accuracy rate in visual assembly?
- GPT-6 Astra achieved an 80 percent accuracy rate in detecting deliberate assembly errors in recent tests.
- Why isn't GPT-6 Astra ready for real-time assistance?
- The model currently takes about three minutes to process each photo, which rules out immediate live feedback for tasks.
Comments
0 comments
Be the first to comment.
Continue Lendo

Meta Muse AI agent: what changes for Brazilian retail
Meta's Muse AI agent brings persistent shopping automation, raising new challenges for Brazilian e-commerce, digital ads, and platform security.

Google Replaces Gemini Gems With Skills in November
Google is replacing Gemini Gems with a new skills system on November 17, 2026, changing how custom AI assistants are invoked.

Claude Sonnet 5.5 Cuts Task Costs by 30% Without Losing Power
Anthropic's Claude Sonnet 5.5 delivers near-Opus performance while cutting task costs by up to 30% through efficient token use.