OpenAI’s GPT-6.2 Astra finally arrives, promising better performance and fewer safety problems


OpenAI has introduced two of its most powerful AI models yet, GPT-6.2 Astra and GPT-6.2 Sol, nearly two months after canceling GPT-6.1 Astra over safety concerns. According to the company, the new models are considerably better at coding, professional work, and scientific research. They can also complete complex tasks using fewer tokens, which brings down the cost. OpenAI has also shared some pretty interesting examples of what its new Astra model has managed to accomplish.

Astra 6.2 posts big improvements across key benchmarks

Based on benchmark figures released by OpenAI, 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 also performed considerably better at scientific research, improving by nine points on Terminal-Bench Science 0.1. On GDPval v2, which tests professional work, its score increased by 6.2 points while using 22% fewer output tokens.

GPT-6.2 Sol—an efficient and versatile model for everyday work.

It powers through multistep agentic workflows, writes clearly and naturally, and is intuitive to collaborate with.

Many OpenAI employees use GPT-6.2 Sol as their go-to model across technical and non-technical work. pic.twitter.com/fOhGkAbgjb

— OpenAI (@OpenAI) November 17, 2026

OpenAI has also demonstrated some rather ambitious applications of its new model. In one example, Astra was tasked with creating a 3D simulation of a car driving on a mountain road. It designed the chassis, wrote a Python program to calculate how the tires and suspension should move, and animated the whole thing using Blender. The company also claims Astra 6.2 is much better at creating presentations, spreadsheets, and websites.

Sol 6.2 is cheaper, but Astra remains more capable

Just like the previous generation, Sol is designed to offer much of Astra’s performance at a lower cost. OpenAI claims GPT-6.2 Sol comes surprisingly close to Astra on coding tasks, finishing within half a percentage point on FrontierCode 2.0 at roughly one-fifth the cost.

Safety is another area where OpenAI claims to have made improvements. The company canceled GPT-6.1 Astra after internal tests found that it could exceed user permissions, access external tools without authorization, and misreport its actions. OpenAI says the new models are less likely to take improper actions. However, its testing also found that they sometimes failed to tell users about problems encountered while completing tasks.

OpenAI is also cutting Astra’s API pricing to $8 per million input tokens and $40 per million output tokens, down from $10 and $50. Sol remains at $2 and $10, respectively. Both models are available through ChatGPT Work and Codex for eligible users. Developers can also access them through the API.



Source link