Breaking

GPT-6 Astra Recreates Lewis Hamilton's Ferrari in Blender After Researching F1 Specs

September 6, 2026 at 18:15 EDT

Foundation Models
AI Agents
Media Generation

An OpenAI employee asked GPT-6 Astra to recreate Lewis Hamilton's Ferrari. The model researched Formula 1 specifications, technical regulations, engineer videos and photographs of the car, then built a full 3D model in Blender with a parts breakdown, plus clay and wireframe versions.

September 5, 2026 · OpenAI

Hamilton's Ferrari, rebuilt in Blender

GPT-6 Astra studied F1 specs, regulations, engineer videos and car photos, then delivered an editable 3D model with a parts breakdown plus clay and wireframe versions.

How the car was built

F1 specs

Public car dimensions and technical data

→

Rules and video

FIA regulations and F1 engineer footage

→

Car photos

Real Hamilton Ferrari reference images

→

Blender model

Parts breakdown, clay and wireframe

Desktop task time

On OSWorld 2.0, Astra finishes ordinary computer tasks in about 40 minutes versus about 75 minutes for GPT-5.6 Sol.

75 min

GPT-5.6 Sol

40 min

GPT-6 Astra

Same unit, drawn to scale: 15 blocks versus 8 blocks (5 minutes each).

72.6%

OSWorld 2.0, versus 65.7% for Sol

95.9%

BenchCAD 3D CAD code, versus 83.3%

50%

Fewer manual fixes in Playco prototyping

$10 / $50

API price per 1M input / output tokens

What testers praise

Developers say the 3D output is inspectable geometry, not a flat render, and that research-then-build in Blender is now strong enough for first-pass game assets.

What critics flag

Some Blender artists argue autonomous modeling sidelines human work. The Ferrari is a visualization, not factory-validated CAD, and outputs can show a generic AI look unless steered.

The limit that still holds

A car you can orbit, not a car you can race

Astra returned clay, wireframe and parts views of Hamilton's Ferrari. It did not produce a scrutineering-legal design or a wind-tunnel model.

Editable Blender meshNo Ferrari validation reportedStill needs human review

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.