FlashNEWS BRIEFING

GPT-6.2 Astra Tops Blueprint-Bench 2 for 3D Spatial Understanding

1 min read

88Z AI, 20 November — OpenAI's GPT-6.2 Astra model secured the top spot on the Blueprint-Bench 2 benchmark, which evaluates AI agents' ability to draw floorplans from photographs of apartment interiors, demonstrating superior understanding of the physical 3D world. The related GPT-6.2 Sol model ranked third.

At a glance

  • GPT-6.2 Astra ranks first on Blueprint-Bench 2, demonstrating leading 3D spatial understanding
  • GPT-6.2 Sol ranks third in the same benchmark
Illustration of GPT-6.2 Astra ranking first on Blueprint-Bench 2
Illustration of GPT-6.2 Astra ranking first on Blueprint-Bench 2

Benchmark Performance

GPT-6.2 Astra achieved the number one rank on Blueprint-Bench 2. This benchmark specifically tests the capability of AI agents to generate floorplans based on photographs of apartment interiors. The result indicates that Astra possesses a superior ability to understand and interpret the physical 3D world compared to other AI systems.

Model Comparison

In the same evaluation, OpenAI's GPT-6.2 Sol model also demonstrated strong performance, securing the third position. This highlights the overall technical strength of the GPT-6.2 series in spatial reasoning and visual understanding tasks, with the Astra variant showing particular excellence in processing three-dimensional space.

Sources & editorial note

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.

  1. @andonlabsSource published

Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.

Related reading

All stories
88Z.ai