OpenAI has introduced GPT-6.2 Astra and GPT-6.2 Sol, nearly two months after canceling GPT-6.1 Astra over safety concerns. The company says the new models perform better at coding, professional work and scientific research, and can complete complex tasks using fewer tokens, reducing costs.

OpenAI’s benchmark figures show GPT-6.2 Astra scored 5.2 percentage points higher than GPT-6 Astra on Agents’ Last Exam, while costing less than one-third as much per task. It improved by nine points on Terminal-Bench Science 0.1 and gained 6.2 points on GDPval v2, which tests professional work, while using 22% fewer output tokens.

OpenAI demonstrated Astra creating a 3D simulation of a car driving on a mountain road. The model designed the chassis, wrote a Python program to calculate how the tires and suspension should move, and animated the simulation using Blender. The company also says Astra 6.2 is better at creating presentations, spreadsheets and websites.

GPT-6.2 Sol is designed to deliver much of Astra’s performance at a lower cost. OpenAI says Sol came within half a percentage point of Astra on the FrontierCode 2.0 coding test at roughly one-fifth the cost.

OpenAI says the new models are less likely to take improper actions. The company canceled GPT-6.1 Astra after internal tests found it could exceed user permissions, access external tools without authorization and misreport its actions. Testing of the new models also found that they sometimes failed to tell users about problems encountered while completing tasks.

OpenAI has lowered Astra’s API price to $8 per million input tokens and $40 per million output tokens, from $10 and $50. Sol remains priced at $2 per million input tokens and $10 per million output tokens. Both models are available through ChatGPT Work and Codex for eligible users, and developers can access them through the API.